Datasets:
Tasks:
Image Classification
Sub-tasks:
multi-class-image-classification
Languages:
English
Size:
1M<n<10M
ArXiv:
License:
Create ForNet.py
Browse files
ForNet.py
ADDED
|
@@ -0,0 +1,1690 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import gzip
|
| 2 |
+
import itertools
|
| 3 |
+
import json
|
| 4 |
+
import multiprocessing
|
| 5 |
+
import os
|
| 6 |
+
import pickle
|
| 7 |
+
import queue
|
| 8 |
+
import re
|
| 9 |
+
import urllib
|
| 10 |
+
import zipfile
|
| 11 |
+
from collections import OrderedDict
|
| 12 |
+
from math import floor
|
| 13 |
+
from typing import Optional
|
| 14 |
+
|
| 15 |
+
import datasets
|
| 16 |
+
import numpy as np
|
| 17 |
+
from datasets import config
|
| 18 |
+
from datasets.arrow_dataset import Dataset
|
| 19 |
+
from datasets.arrow_reader import ArrowReader
|
| 20 |
+
from datasets.fingerprint import Hasher
|
| 21 |
+
from PIL import ImageFilter
|
| 22 |
+
from torchvision import transforms as T
|
| 23 |
+
from tqdm import tqdm
|
| 24 |
+
|
| 25 |
+
logger = datasets.logging.get_logger(__name__)
|
| 26 |
+
|
| 27 |
+
|
| 28 |
+
# taken from https://huggingface.co/datasets/ILSVRC/imagenet-1k/blob/main/classes.py
|
| 29 |
+
IMAGENET2012_CLASSES = OrderedDict(
|
| 30 |
+
{
|
| 31 |
+
"n01440764": "tench, Tinca tinca",
|
| 32 |
+
"n01443537": "goldfish, Carassius auratus",
|
| 33 |
+
"n01484850": "great white shark, white shark, man-eater, man-eating shark, Carcharodon carcharias",
|
| 34 |
+
"n01491361": "tiger shark, Galeocerdo cuvieri",
|
| 35 |
+
"n01494475": "hammerhead, hammerhead shark",
|
| 36 |
+
"n01496331": "electric ray, crampfish, numbfish, torpedo",
|
| 37 |
+
"n01498041": "stingray",
|
| 38 |
+
"n01514668": "cock",
|
| 39 |
+
"n01514859": "hen",
|
| 40 |
+
"n01518878": "ostrich, Struthio camelus",
|
| 41 |
+
"n01530575": "brambling, Fringilla montifringilla",
|
| 42 |
+
"n01531178": "goldfinch, Carduelis carduelis",
|
| 43 |
+
"n01532829": "house finch, linnet, Carpodacus mexicanus",
|
| 44 |
+
"n01534433": "junco, snowbird",
|
| 45 |
+
"n01537544": "indigo bunting, indigo finch, indigo bird, Passerina cyanea",
|
| 46 |
+
"n01558993": "robin, American robin, Turdus migratorius",
|
| 47 |
+
"n01560419": "bulbul",
|
| 48 |
+
"n01580077": "jay",
|
| 49 |
+
"n01582220": "magpie",
|
| 50 |
+
"n01592084": "chickadee",
|
| 51 |
+
"n01601694": "water ouzel, dipper",
|
| 52 |
+
"n01608432": "kite",
|
| 53 |
+
"n01614925": "bald eagle, American eagle, Haliaeetus leucocephalus",
|
| 54 |
+
"n01616318": "vulture",
|
| 55 |
+
"n01622779": "great grey owl, great gray owl, Strix nebulosa",
|
| 56 |
+
"n01629819": "European fire salamander, Salamandra salamandra",
|
| 57 |
+
"n01630670": "common newt, Triturus vulgaris",
|
| 58 |
+
"n01631663": "eft",
|
| 59 |
+
"n01632458": "spotted salamander, Ambystoma maculatum",
|
| 60 |
+
"n01632777": "axolotl, mud puppy, Ambystoma mexicanum",
|
| 61 |
+
"n01641577": "bullfrog, Rana catesbeiana",
|
| 62 |
+
"n01644373": "tree frog, tree-frog",
|
| 63 |
+
"n01644900": "tailed frog, bell toad, ribbed toad, tailed toad, Ascaphus trui",
|
| 64 |
+
"n01664065": "loggerhead, loggerhead turtle, Caretta caretta",
|
| 65 |
+
"n01665541": "leatherback turtle, leatherback, leathery turtle, Dermochelys coriacea",
|
| 66 |
+
"n01667114": "mud turtle",
|
| 67 |
+
"n01667778": "terrapin",
|
| 68 |
+
"n01669191": "box turtle, box tortoise",
|
| 69 |
+
"n01675722": "banded gecko",
|
| 70 |
+
"n01677366": "common iguana, iguana, Iguana iguana",
|
| 71 |
+
"n01682714": "American chameleon, anole, Anolis carolinensis",
|
| 72 |
+
"n01685808": "whiptail, whiptail lizard",
|
| 73 |
+
"n01687978": "agama",
|
| 74 |
+
"n01688243": "frilled lizard, Chlamydosaurus kingi",
|
| 75 |
+
"n01689811": "alligator lizard",
|
| 76 |
+
"n01692333": "Gila monster, Heloderma suspectum",
|
| 77 |
+
"n01693334": "green lizard, Lacerta viridis",
|
| 78 |
+
"n01694178": "African chameleon, Chamaeleo chamaeleon",
|
| 79 |
+
"n01695060": "Komodo dragon, Komodo lizard, dragon lizard, giant lizard, Varanus komodoensis",
|
| 80 |
+
"n01697457": "African crocodile, Nile crocodile, Crocodylus niloticus",
|
| 81 |
+
"n01698640": "American alligator, Alligator mississipiensis",
|
| 82 |
+
"n01704323": "triceratops",
|
| 83 |
+
"n01728572": "thunder snake, worm snake, Carphophis amoenus",
|
| 84 |
+
"n01728920": "ringneck snake, ring-necked snake, ring snake",
|
| 85 |
+
"n01729322": "hognose snake, puff adder, sand viper",
|
| 86 |
+
"n01729977": "green snake, grass snake",
|
| 87 |
+
"n01734418": "king snake, kingsnake",
|
| 88 |
+
"n01735189": "garter snake, grass snake",
|
| 89 |
+
"n01737021": "water snake",
|
| 90 |
+
"n01739381": "vine snake",
|
| 91 |
+
"n01740131": "night snake, Hypsiglena torquata",
|
| 92 |
+
"n01742172": "boa constrictor, Constrictor constrictor",
|
| 93 |
+
"n01744401": "rock python, rock snake, Python sebae",
|
| 94 |
+
"n01748264": "Indian cobra, Naja naja",
|
| 95 |
+
"n01749939": "green mamba",
|
| 96 |
+
"n01751748": "sea snake",
|
| 97 |
+
"n01753488": "horned viper, cerastes, sand viper, horned asp, Cerastes cornutus",
|
| 98 |
+
"n01755581": "diamondback, diamondback rattlesnake, Crotalus adamanteus",
|
| 99 |
+
"n01756291": "sidewinder, horned rattlesnake, Crotalus cerastes",
|
| 100 |
+
"n01768244": "trilobite",
|
| 101 |
+
"n01770081": "harvestman, daddy longlegs, Phalangium opilio",
|
| 102 |
+
"n01770393": "scorpion",
|
| 103 |
+
"n01773157": "black and gold garden spider, Argiope aurantia",
|
| 104 |
+
"n01773549": "barn spider, Araneus cavaticus",
|
| 105 |
+
"n01773797": "garden spider, Aranea diademata",
|
| 106 |
+
"n01774384": "black widow, Latrodectus mactans",
|
| 107 |
+
"n01774750": "tarantula",
|
| 108 |
+
"n01775062": "wolf spider, hunting spider",
|
| 109 |
+
"n01776313": "tick",
|
| 110 |
+
"n01784675": "centipede",
|
| 111 |
+
"n01795545": "black grouse",
|
| 112 |
+
"n01796340": "ptarmigan",
|
| 113 |
+
"n01797886": "ruffed grouse, partridge, Bonasa umbellus",
|
| 114 |
+
"n01798484": "prairie chicken, prairie grouse, prairie fowl",
|
| 115 |
+
"n01806143": "peacock",
|
| 116 |
+
"n01806567": "quail",
|
| 117 |
+
"n01807496": "partridge",
|
| 118 |
+
"n01817953": "African grey, African gray, Psittacus erithacus",
|
| 119 |
+
"n01818515": "macaw",
|
| 120 |
+
"n01819313": "sulphur-crested cockatoo, Kakatoe galerita, Cacatua galerita",
|
| 121 |
+
"n01820546": "lorikeet",
|
| 122 |
+
"n01824575": "coucal",
|
| 123 |
+
"n01828970": "bee eater",
|
| 124 |
+
"n01829413": "hornbill",
|
| 125 |
+
"n01833805": "hummingbird",
|
| 126 |
+
"n01843065": "jacamar",
|
| 127 |
+
"n01843383": "toucan",
|
| 128 |
+
"n01847000": "drake",
|
| 129 |
+
"n01855032": "red-breasted merganser, Mergus serrator",
|
| 130 |
+
"n01855672": "goose",
|
| 131 |
+
"n01860187": "black swan, Cygnus atratus",
|
| 132 |
+
"n01871265": "tusker",
|
| 133 |
+
"n01872401": "echidna, spiny anteater, anteater",
|
| 134 |
+
"n01873310": "platypus, duckbill, duckbilled platypus, duck-billed platypus, Ornithorhynchus anatinus",
|
| 135 |
+
"n01877812": "wallaby, brush kangaroo",
|
| 136 |
+
"n01882714": "koala, koala bear, kangaroo bear, native bear, Phascolarctos cinereus",
|
| 137 |
+
"n01883070": "wombat",
|
| 138 |
+
"n01910747": "jellyfish",
|
| 139 |
+
"n01914609": "sea anemone, anemone",
|
| 140 |
+
"n01917289": "brain coral",
|
| 141 |
+
"n01924916": "flatworm, platyhelminth",
|
| 142 |
+
"n01930112": "nematode, nematode worm, roundworm",
|
| 143 |
+
"n01943899": "conch",
|
| 144 |
+
"n01944390": "snail",
|
| 145 |
+
"n01945685": "slug",
|
| 146 |
+
"n01950731": "sea slug, nudibranch",
|
| 147 |
+
"n01955084": "chiton, coat-of-mail shell, sea cradle, polyplacophore",
|
| 148 |
+
"n01968897": "chambered nautilus, pearly nautilus, nautilus",
|
| 149 |
+
"n01978287": "Dungeness crab, Cancer magister",
|
| 150 |
+
"n01978455": "rock crab, Cancer irroratus",
|
| 151 |
+
"n01980166": "fiddler crab",
|
| 152 |
+
"n01981276": "king crab, Alaska crab, Alaskan king crab, Alaska king crab, Paralithodes camtschatica",
|
| 153 |
+
"n01983481": "American lobster, Northern lobster, Maine lobster, Homarus americanus",
|
| 154 |
+
"n01984695": "spiny lobster, langouste, rock lobster, crawfish, crayfish, sea crawfish",
|
| 155 |
+
"n01985128": "crayfish, crawfish, crawdad, crawdaddy",
|
| 156 |
+
"n01986214": "hermit crab",
|
| 157 |
+
"n01990800": "isopod",
|
| 158 |
+
"n02002556": "white stork, Ciconia ciconia",
|
| 159 |
+
"n02002724": "black stork, Ciconia nigra",
|
| 160 |
+
"n02006656": "spoonbill",
|
| 161 |
+
"n02007558": "flamingo",
|
| 162 |
+
"n02009229": "little blue heron, Egretta caerulea",
|
| 163 |
+
"n02009912": "American egret, great white heron, Egretta albus",
|
| 164 |
+
"n02011460": "bittern",
|
| 165 |
+
"n02012849": "crane",
|
| 166 |
+
"n02013706": "limpkin, Aramus pictus",
|
| 167 |
+
"n02017213": "European gallinule, Porphyrio porphyrio",
|
| 168 |
+
"n02018207": "American coot, marsh hen, mud hen, water hen, Fulica americana",
|
| 169 |
+
"n02018795": "bustard",
|
| 170 |
+
"n02025239": "ruddy turnstone, Arenaria interpres",
|
| 171 |
+
"n02027492": "red-backed sandpiper, dunlin, Erolia alpina",
|
| 172 |
+
"n02028035": "redshank, Tringa totanus",
|
| 173 |
+
"n02033041": "dowitcher",
|
| 174 |
+
"n02037110": "oystercatcher, oyster catcher",
|
| 175 |
+
"n02051845": "pelican",
|
| 176 |
+
"n02056570": "king penguin, Aptenodytes patagonica",
|
| 177 |
+
"n02058221": "albatross, mollymawk",
|
| 178 |
+
"n02066245": "grey whale, gray whale, devilfish, Eschrichtius gibbosus, Eschrichtius robustus",
|
| 179 |
+
"n02071294": "killer whale, killer, orca, grampus, sea wolf, Orcinus orca",
|
| 180 |
+
"n02074367": "dugong, Dugong dugon",
|
| 181 |
+
"n02077923": "sea lion",
|
| 182 |
+
"n02085620": "Chihuahua",
|
| 183 |
+
"n02085782": "Japanese spaniel",
|
| 184 |
+
"n02085936": "Maltese dog, Maltese terrier, Maltese",
|
| 185 |
+
"n02086079": "Pekinese, Pekingese, Peke",
|
| 186 |
+
"n02086240": "Shih-Tzu",
|
| 187 |
+
"n02086646": "Blenheim spaniel",
|
| 188 |
+
"n02086910": "papillon",
|
| 189 |
+
"n02087046": "toy terrier",
|
| 190 |
+
"n02087394": "Rhodesian ridgeback",
|
| 191 |
+
"n02088094": "Afghan hound, Afghan",
|
| 192 |
+
"n02088238": "basset, basset hound",
|
| 193 |
+
"n02088364": "beagle",
|
| 194 |
+
"n02088466": "bloodhound, sleuthhound",
|
| 195 |
+
"n02088632": "bluetick",
|
| 196 |
+
"n02089078": "black-and-tan coonhound",
|
| 197 |
+
"n02089867": "Walker hound, Walker foxhound",
|
| 198 |
+
"n02089973": "English foxhound",
|
| 199 |
+
"n02090379": "redbone",
|
| 200 |
+
"n02090622": "borzoi, Russian wolfhound",
|
| 201 |
+
"n02090721": "Irish wolfhound",
|
| 202 |
+
"n02091032": "Italian greyhound",
|
| 203 |
+
"n02091134": "whippet",
|
| 204 |
+
"n02091244": "Ibizan hound, Ibizan Podenco",
|
| 205 |
+
"n02091467": "Norwegian elkhound, elkhound",
|
| 206 |
+
"n02091635": "otterhound, otter hound",
|
| 207 |
+
"n02091831": "Saluki, gazelle hound",
|
| 208 |
+
"n02092002": "Scottish deerhound, deerhound",
|
| 209 |
+
"n02092339": "Weimaraner",
|
| 210 |
+
"n02093256": "Staffordshire bullterrier, Staffordshire bull terrier",
|
| 211 |
+
"n02093428": "American Staffordshire terrier, Staffordshire terrier, American pit bull terrier, pit bull terrier",
|
| 212 |
+
"n02093647": "Bedlington terrier",
|
| 213 |
+
"n02093754": "Border terrier",
|
| 214 |
+
"n02093859": "Kerry blue terrier",
|
| 215 |
+
"n02093991": "Irish terrier",
|
| 216 |
+
"n02094114": "Norfolk terrier",
|
| 217 |
+
"n02094258": "Norwich terrier",
|
| 218 |
+
"n02094433": "Yorkshire terrier",
|
| 219 |
+
"n02095314": "wire-haired fox terrier",
|
| 220 |
+
"n02095570": "Lakeland terrier",
|
| 221 |
+
"n02095889": "Sealyham terrier, Sealyham",
|
| 222 |
+
"n02096051": "Airedale, Airedale terrier",
|
| 223 |
+
"n02096177": "cairn, cairn terrier",
|
| 224 |
+
"n02096294": "Australian terrier",
|
| 225 |
+
"n02096437": "Dandie Dinmont, Dandie Dinmont terrier",
|
| 226 |
+
"n02096585": "Boston bull, Boston terrier",
|
| 227 |
+
"n02097047": "miniature schnauzer",
|
| 228 |
+
"n02097130": "giant schnauzer",
|
| 229 |
+
"n02097209": "standard schnauzer",
|
| 230 |
+
"n02097298": "Scotch terrier, Scottish terrier, Scottie",
|
| 231 |
+
"n02097474": "Tibetan terrier, chrysanthemum dog",
|
| 232 |
+
"n02097658": "silky terrier, Sydney silky",
|
| 233 |
+
"n02098105": "soft-coated wheaten terrier",
|
| 234 |
+
"n02098286": "West Highland white terrier",
|
| 235 |
+
"n02098413": "Lhasa, Lhasa apso",
|
| 236 |
+
"n02099267": "flat-coated retriever",
|
| 237 |
+
"n02099429": "curly-coated retriever",
|
| 238 |
+
"n02099601": "golden retriever",
|
| 239 |
+
"n02099712": "Labrador retriever",
|
| 240 |
+
"n02099849": "Chesapeake Bay retriever",
|
| 241 |
+
"n02100236": "German short-haired pointer",
|
| 242 |
+
"n02100583": "vizsla, Hungarian pointer",
|
| 243 |
+
"n02100735": "English setter",
|
| 244 |
+
"n02100877": "Irish setter, red setter",
|
| 245 |
+
"n02101006": "Gordon setter",
|
| 246 |
+
"n02101388": "Brittany spaniel",
|
| 247 |
+
"n02101556": "clumber, clumber spaniel",
|
| 248 |
+
"n02102040": "English springer, English springer spaniel",
|
| 249 |
+
"n02102177": "Welsh springer spaniel",
|
| 250 |
+
"n02102318": "cocker spaniel, English cocker spaniel, cocker",
|
| 251 |
+
"n02102480": "Sussex spaniel",
|
| 252 |
+
"n02102973": "Irish water spaniel",
|
| 253 |
+
"n02104029": "kuvasz",
|
| 254 |
+
"n02104365": "schipperke",
|
| 255 |
+
"n02105056": "groenendael",
|
| 256 |
+
"n02105162": "malinois",
|
| 257 |
+
"n02105251": "briard",
|
| 258 |
+
"n02105412": "kelpie",
|
| 259 |
+
"n02105505": "komondor",
|
| 260 |
+
"n02105641": "Old English sheepdog, bobtail",
|
| 261 |
+
"n02105855": "Shetland sheepdog, Shetland sheep dog, Shetland",
|
| 262 |
+
"n02106030": "collie",
|
| 263 |
+
"n02106166": "Border collie",
|
| 264 |
+
"n02106382": "Bouvier des Flandres, Bouviers des Flandres",
|
| 265 |
+
"n02106550": "Rottweiler",
|
| 266 |
+
"n02106662": "German shepherd, German shepherd dog, German police dog, alsatian",
|
| 267 |
+
"n02107142": "Doberman, Doberman pinscher",
|
| 268 |
+
"n02107312": "miniature pinscher",
|
| 269 |
+
"n02107574": "Greater Swiss Mountain dog",
|
| 270 |
+
"n02107683": "Bernese mountain dog",
|
| 271 |
+
"n02107908": "Appenzeller",
|
| 272 |
+
"n02108000": "EntleBucher",
|
| 273 |
+
"n02108089": "boxer",
|
| 274 |
+
"n02108422": "bull mastiff",
|
| 275 |
+
"n02108551": "Tibetan mastiff",
|
| 276 |
+
"n02108915": "French bulldog",
|
| 277 |
+
"n02109047": "Great Dane",
|
| 278 |
+
"n02109525": "Saint Bernard, St Bernard",
|
| 279 |
+
"n02109961": "Eskimo dog, husky",
|
| 280 |
+
"n02110063": "malamute, malemute, Alaskan malamute",
|
| 281 |
+
"n02110185": "Siberian husky",
|
| 282 |
+
"n02110341": "dalmatian, coach dog, carriage dog",
|
| 283 |
+
"n02110627": "affenpinscher, monkey pinscher, monkey dog",
|
| 284 |
+
"n02110806": "basenji",
|
| 285 |
+
"n02110958": "pug, pug-dog",
|
| 286 |
+
"n02111129": "Leonberg",
|
| 287 |
+
"n02111277": "Newfoundland, Newfoundland dog",
|
| 288 |
+
"n02111500": "Great Pyrenees",
|
| 289 |
+
"n02111889": "Samoyed, Samoyede",
|
| 290 |
+
"n02112018": "Pomeranian",
|
| 291 |
+
"n02112137": "chow, chow chow",
|
| 292 |
+
"n02112350": "keeshond",
|
| 293 |
+
"n02112706": "Brabancon griffon",
|
| 294 |
+
"n02113023": "Pembroke, Pembroke Welsh corgi",
|
| 295 |
+
"n02113186": "Cardigan, Cardigan Welsh corgi",
|
| 296 |
+
"n02113624": "toy poodle",
|
| 297 |
+
"n02113712": "miniature poodle",
|
| 298 |
+
"n02113799": "standard poodle",
|
| 299 |
+
"n02113978": "Mexican hairless",
|
| 300 |
+
"n02114367": "timber wolf, grey wolf, gray wolf, Canis lupus",
|
| 301 |
+
"n02114548": "white wolf, Arctic wolf, Canis lupus tundrarum",
|
| 302 |
+
"n02114712": "red wolf, maned wolf, Canis rufus, Canis niger",
|
| 303 |
+
"n02114855": "coyote, prairie wolf, brush wolf, Canis latrans",
|
| 304 |
+
"n02115641": "dingo, warrigal, warragal, Canis dingo",
|
| 305 |
+
"n02115913": "dhole, Cuon alpinus",
|
| 306 |
+
"n02116738": "African hunting dog, hyena dog, Cape hunting dog, Lycaon pictus",
|
| 307 |
+
"n02117135": "hyena, hyaena",
|
| 308 |
+
"n02119022": "red fox, Vulpes vulpes",
|
| 309 |
+
"n02119789": "kit fox, Vulpes macrotis",
|
| 310 |
+
"n02120079": "Arctic fox, white fox, Alopex lagopus",
|
| 311 |
+
"n02120505": "grey fox, gray fox, Urocyon cinereoargenteus",
|
| 312 |
+
"n02123045": "tabby, tabby cat",
|
| 313 |
+
"n02123159": "tiger cat",
|
| 314 |
+
"n02123394": "Persian cat",
|
| 315 |
+
"n02123597": "Siamese cat, Siamese",
|
| 316 |
+
"n02124075": "Egyptian cat",
|
| 317 |
+
"n02125311": "cougar, puma, catamount, mountain lion, painter, panther, Felis concolor",
|
| 318 |
+
"n02127052": "lynx, catamount",
|
| 319 |
+
"n02128385": "leopard, Panthera pardus",
|
| 320 |
+
"n02128757": "snow leopard, ounce, Panthera uncia",
|
| 321 |
+
"n02128925": "jaguar, panther, Panthera onca, Felis onca",
|
| 322 |
+
"n02129165": "lion, king of beasts, Panthera leo",
|
| 323 |
+
"n02129604": "tiger, Panthera tigris",
|
| 324 |
+
"n02130308": "cheetah, chetah, Acinonyx jubatus",
|
| 325 |
+
"n02132136": "brown bear, bruin, Ursus arctos",
|
| 326 |
+
"n02133161": "American black bear, black bear, Ursus americanus, Euarctos americanus",
|
| 327 |
+
"n02134084": "ice bear, polar bear, Ursus Maritimus, Thalarctos maritimus",
|
| 328 |
+
"n02134418": "sloth bear, Melursus ursinus, Ursus ursinus",
|
| 329 |
+
"n02137549": "mongoose",
|
| 330 |
+
"n02138441": "meerkat, mierkat",
|
| 331 |
+
"n02165105": "tiger beetle",
|
| 332 |
+
"n02165456": "ladybug, ladybeetle, lady beetle, ladybird, ladybird beetle",
|
| 333 |
+
"n02167151": "ground beetle, carabid beetle",
|
| 334 |
+
"n02168699": "long-horned beetle, longicorn, longicorn beetle",
|
| 335 |
+
"n02169497": "leaf beetle, chrysomelid",
|
| 336 |
+
"n02172182": "dung beetle",
|
| 337 |
+
"n02174001": "rhinoceros beetle",
|
| 338 |
+
"n02177972": "weevil",
|
| 339 |
+
"n02190166": "fly",
|
| 340 |
+
"n02206856": "bee",
|
| 341 |
+
"n02219486": "ant, emmet, pismire",
|
| 342 |
+
"n02226429": "grasshopper, hopper",
|
| 343 |
+
"n02229544": "cricket",
|
| 344 |
+
"n02231487": "walking stick, walkingstick, stick insect",
|
| 345 |
+
"n02233338": "cockroach, roach",
|
| 346 |
+
"n02236044": "mantis, mantid",
|
| 347 |
+
"n02256656": "cicada, cicala",
|
| 348 |
+
"n02259212": "leafhopper",
|
| 349 |
+
"n02264363": "lacewing, lacewing fly",
|
| 350 |
+
"n02268443": "dragonfly, darning needle, devil's darning needle, sewing needle, snake feeder, snake doctor, mosquito hawk, skeeter hawk",
|
| 351 |
+
"n02268853": "damselfly",
|
| 352 |
+
"n02276258": "admiral",
|
| 353 |
+
"n02277742": "ringlet, ringlet butterfly",
|
| 354 |
+
"n02279972": "monarch, monarch butterfly, milkweed butterfly, Danaus plexippus",
|
| 355 |
+
"n02280649": "cabbage butterfly",
|
| 356 |
+
"n02281406": "sulphur butterfly, sulfur butterfly",
|
| 357 |
+
"n02281787": "lycaenid, lycaenid butterfly",
|
| 358 |
+
"n02317335": "starfish, sea star",
|
| 359 |
+
"n02319095": "sea urchin",
|
| 360 |
+
"n02321529": "sea cucumber, holothurian",
|
| 361 |
+
"n02325366": "wood rabbit, cottontail, cottontail rabbit",
|
| 362 |
+
"n02326432": "hare",
|
| 363 |
+
"n02328150": "Angora, Angora rabbit",
|
| 364 |
+
"n02342885": "hamster",
|
| 365 |
+
"n02346627": "porcupine, hedgehog",
|
| 366 |
+
"n02356798": "fox squirrel, eastern fox squirrel, Sciurus niger",
|
| 367 |
+
"n02361337": "marmot",
|
| 368 |
+
"n02363005": "beaver",
|
| 369 |
+
"n02364673": "guinea pig, Cavia cobaya",
|
| 370 |
+
"n02389026": "sorrel",
|
| 371 |
+
"n02391049": "zebra",
|
| 372 |
+
"n02395406": "hog, pig, grunter, squealer, Sus scrofa",
|
| 373 |
+
"n02396427": "wild boar, boar, Sus scrofa",
|
| 374 |
+
"n02397096": "warthog",
|
| 375 |
+
"n02398521": "hippopotamus, hippo, river horse, Hippopotamus amphibius",
|
| 376 |
+
"n02403003": "ox",
|
| 377 |
+
"n02408429": "water buffalo, water ox, Asiatic buffalo, Bubalus bubalis",
|
| 378 |
+
"n02410509": "bison",
|
| 379 |
+
"n02412080": "ram, tup",
|
| 380 |
+
"n02415577": "bighorn, bighorn sheep, cimarron, Rocky Mountain bighorn, Rocky Mountain sheep, Ovis canadensis",
|
| 381 |
+
"n02417914": "ibex, Capra ibex",
|
| 382 |
+
"n02422106": "hartebeest",
|
| 383 |
+
"n02422699": "impala, Aepyceros melampus",
|
| 384 |
+
"n02423022": "gazelle",
|
| 385 |
+
"n02437312": "Arabian camel, dromedary, Camelus dromedarius",
|
| 386 |
+
"n02437616": "llama",
|
| 387 |
+
"n02441942": "weasel",
|
| 388 |
+
"n02442845": "mink",
|
| 389 |
+
"n02443114": "polecat, fitch, foulmart, foumart, Mustela putorius",
|
| 390 |
+
"n02443484": "black-footed ferret, ferret, Mustela nigripes",
|
| 391 |
+
"n02444819": "otter",
|
| 392 |
+
"n02445715": "skunk, polecat, wood pussy",
|
| 393 |
+
"n02447366": "badger",
|
| 394 |
+
"n02454379": "armadillo",
|
| 395 |
+
"n02457408": "three-toed sloth, ai, Bradypus tridactylus",
|
| 396 |
+
"n02480495": "orangutan, orang, orangutang, Pongo pygmaeus",
|
| 397 |
+
"n02480855": "gorilla, Gorilla gorilla",
|
| 398 |
+
"n02481823": "chimpanzee, chimp, Pan troglodytes",
|
| 399 |
+
"n02483362": "gibbon, Hylobates lar",
|
| 400 |
+
"n02483708": "siamang, Hylobates syndactylus, Symphalangus syndactylus",
|
| 401 |
+
"n02484975": "guenon, guenon monkey",
|
| 402 |
+
"n02486261": "patas, hussar monkey, Erythrocebus patas",
|
| 403 |
+
"n02486410": "baboon",
|
| 404 |
+
"n02487347": "macaque",
|
| 405 |
+
"n02488291": "langur",
|
| 406 |
+
"n02488702": "colobus, colobus monkey",
|
| 407 |
+
"n02489166": "proboscis monkey, Nasalis larvatus",
|
| 408 |
+
"n02490219": "marmoset",
|
| 409 |
+
"n02492035": "capuchin, ringtail, Cebus capucinus",
|
| 410 |
+
"n02492660": "howler monkey, howler",
|
| 411 |
+
"n02493509": "titi, titi monkey",
|
| 412 |
+
"n02493793": "spider monkey, Ateles geoffroyi",
|
| 413 |
+
"n02494079": "squirrel monkey, Saimiri sciureus",
|
| 414 |
+
"n02497673": "Madagascar cat, ring-tailed lemur, Lemur catta",
|
| 415 |
+
"n02500267": "indri, indris, Indri indri, Indri brevicaudatus",
|
| 416 |
+
"n02504013": "Indian elephant, Elephas maximus",
|
| 417 |
+
"n02504458": "African elephant, Loxodonta africana",
|
| 418 |
+
"n02509815": "lesser panda, red panda, panda, bear cat, cat bear, Ailurus fulgens",
|
| 419 |
+
"n02510455": "giant panda, panda, panda bear, coon bear, Ailuropoda melanoleuca",
|
| 420 |
+
"n02514041": "barracouta, snoek",
|
| 421 |
+
"n02526121": "eel",
|
| 422 |
+
"n02536864": "coho, cohoe, coho salmon, blue jack, silver salmon, Oncorhynchus kisutch",
|
| 423 |
+
"n02606052": "rock beauty, Holocanthus tricolor",
|
| 424 |
+
"n02607072": "anemone fish",
|
| 425 |
+
"n02640242": "sturgeon",
|
| 426 |
+
"n02641379": "gar, garfish, garpike, billfish, Lepisosteus osseus",
|
| 427 |
+
"n02643566": "lionfish",
|
| 428 |
+
"n02655020": "puffer, pufferfish, blowfish, globefish",
|
| 429 |
+
"n02666196": "abacus",
|
| 430 |
+
"n02667093": "abaya",
|
| 431 |
+
"n02669723": "academic gown, academic robe, judge's robe",
|
| 432 |
+
"n02672831": "accordion, piano accordion, squeeze box",
|
| 433 |
+
"n02676566": "acoustic guitar",
|
| 434 |
+
"n02687172": "aircraft carrier, carrier, flattop, attack aircraft carrier",
|
| 435 |
+
"n02690373": "airliner",
|
| 436 |
+
"n02692877": "airship, dirigible",
|
| 437 |
+
"n02699494": "altar",
|
| 438 |
+
"n02701002": "ambulance",
|
| 439 |
+
"n02704792": "amphibian, amphibious vehicle",
|
| 440 |
+
"n02708093": "analog clock",
|
| 441 |
+
"n02727426": "apiary, bee house",
|
| 442 |
+
"n02730930": "apron",
|
| 443 |
+
"n02747177": "ashcan, trash can, garbage can, wastebin, ash bin, ash-bin, ashbin, dustbin, trash barrel, trash bin",
|
| 444 |
+
"n02749479": "assault rifle, assault gun",
|
| 445 |
+
"n02769748": "backpack, back pack, knapsack, packsack, rucksack, haversack",
|
| 446 |
+
"n02776631": "bakery, bakeshop, bakehouse",
|
| 447 |
+
"n02777292": "balance beam, beam",
|
| 448 |
+
"n02782093": "balloon",
|
| 449 |
+
"n02783161": "ballpoint, ballpoint pen, ballpen, Biro",
|
| 450 |
+
"n02786058": "Band Aid",
|
| 451 |
+
"n02787622": "banjo",
|
| 452 |
+
"n02788148": "bannister, banister, balustrade, balusters, handrail",
|
| 453 |
+
"n02790996": "barbell",
|
| 454 |
+
"n02791124": "barber chair",
|
| 455 |
+
"n02791270": "barbershop",
|
| 456 |
+
"n02793495": "barn",
|
| 457 |
+
"n02794156": "barometer",
|
| 458 |
+
"n02795169": "barrel, cask",
|
| 459 |
+
"n02797295": "barrow, garden cart, lawn cart, wheelbarrow",
|
| 460 |
+
"n02799071": "baseball",
|
| 461 |
+
"n02802426": "basketball",
|
| 462 |
+
"n02804414": "bassinet",
|
| 463 |
+
"n02804610": "bassoon",
|
| 464 |
+
"n02807133": "bathing cap, swimming cap",
|
| 465 |
+
"n02808304": "bath towel",
|
| 466 |
+
"n02808440": "bathtub, bathing tub, bath, tub",
|
| 467 |
+
"n02814533": "beach wagon, station wagon, wagon, estate car, beach waggon, station waggon, waggon",
|
| 468 |
+
"n02814860": "beacon, lighthouse, beacon light, pharos",
|
| 469 |
+
"n02815834": "beaker",
|
| 470 |
+
"n02817516": "bearskin, busby, shako",
|
| 471 |
+
"n02823428": "beer bottle",
|
| 472 |
+
"n02823750": "beer glass",
|
| 473 |
+
"n02825657": "bell cote, bell cot",
|
| 474 |
+
"n02834397": "bib",
|
| 475 |
+
"n02835271": "bicycle-built-for-two, tandem bicycle, tandem",
|
| 476 |
+
"n02837789": "bikini, two-piece",
|
| 477 |
+
"n02840245": "binder, ring-binder",
|
| 478 |
+
"n02841315": "binoculars, field glasses, opera glasses",
|
| 479 |
+
"n02843684": "birdhouse",
|
| 480 |
+
"n02859443": "boathouse",
|
| 481 |
+
"n02860847": "bobsled, bobsleigh, bob",
|
| 482 |
+
"n02865351": "bolo tie, bolo, bola tie, bola",
|
| 483 |
+
"n02869837": "bonnet, poke bonnet",
|
| 484 |
+
"n02870880": "bookcase",
|
| 485 |
+
"n02871525": "bookshop, bookstore, bookstall",
|
| 486 |
+
"n02877765": "bottlecap",
|
| 487 |
+
"n02879718": "bow",
|
| 488 |
+
"n02883205": "bow tie, bow-tie, bowtie",
|
| 489 |
+
"n02892201": "brass, memorial tablet, plaque",
|
| 490 |
+
"n02892767": "brassiere, bra, bandeau",
|
| 491 |
+
"n02894605": "breakwater, groin, groyne, mole, bulwark, seawall, jetty",
|
| 492 |
+
"n02895154": "breastplate, aegis, egis",
|
| 493 |
+
"n02906734": "broom",
|
| 494 |
+
"n02909870": "bucket, pail",
|
| 495 |
+
"n02910353": "buckle",
|
| 496 |
+
"n02916936": "bulletproof vest",
|
| 497 |
+
"n02917067": "bullet train, bullet",
|
| 498 |
+
"n02927161": "butcher shop, meat market",
|
| 499 |
+
"n02930766": "cab, hack, taxi, taxicab",
|
| 500 |
+
"n02939185": "caldron, cauldron",
|
| 501 |
+
"n02948072": "candle, taper, wax light",
|
| 502 |
+
"n02950826": "cannon",
|
| 503 |
+
"n02951358": "canoe",
|
| 504 |
+
"n02951585": "can opener, tin opener",
|
| 505 |
+
"n02963159": "cardigan",
|
| 506 |
+
"n02965783": "car mirror",
|
| 507 |
+
"n02966193": "carousel, carrousel, merry-go-round, roundabout, whirligig",
|
| 508 |
+
"n02966687": "carpenter's kit, tool kit",
|
| 509 |
+
"n02971356": "carton",
|
| 510 |
+
"n02974003": "car wheel",
|
| 511 |
+
"n02977058": "cash machine, cash dispenser, automated teller machine, automatic teller machine, automated teller, automatic teller, ATM",
|
| 512 |
+
"n02978881": "cassette",
|
| 513 |
+
"n02979186": "cassette player",
|
| 514 |
+
"n02980441": "castle",
|
| 515 |
+
"n02981792": "catamaran",
|
| 516 |
+
"n02988304": "CD player",
|
| 517 |
+
"n02992211": "cello, violoncello",
|
| 518 |
+
"n02992529": "cellular telephone, cellular phone, cellphone, cell, mobile phone",
|
| 519 |
+
"n02999410": "chain",
|
| 520 |
+
"n03000134": "chainlink fence",
|
| 521 |
+
"n03000247": "chain mail, ring mail, mail, chain armor, chain armour, ring armor, ring armour",
|
| 522 |
+
"n03000684": "chain saw, chainsaw",
|
| 523 |
+
"n03014705": "chest",
|
| 524 |
+
"n03016953": "chiffonier, commode",
|
| 525 |
+
"n03017168": "chime, bell, gong",
|
| 526 |
+
"n03018349": "china cabinet, china closet",
|
| 527 |
+
"n03026506": "Christmas stocking",
|
| 528 |
+
"n03028079": "church, church building",
|
| 529 |
+
"n03032252": "cinema, movie theater, movie theatre, movie house, picture palace",
|
| 530 |
+
"n03041632": "cleaver, meat cleaver, chopper",
|
| 531 |
+
"n03042490": "cliff dwelling",
|
| 532 |
+
"n03045698": "cloak",
|
| 533 |
+
"n03047690": "clog, geta, patten, sabot",
|
| 534 |
+
"n03062245": "cocktail shaker",
|
| 535 |
+
"n03063599": "coffee mug",
|
| 536 |
+
"n03063689": "coffeepot",
|
| 537 |
+
"n03065424": "coil, spiral, volute, whorl, helix",
|
| 538 |
+
"n03075370": "combination lock",
|
| 539 |
+
"n03085013": "computer keyboard, keypad",
|
| 540 |
+
"n03089624": "confectionery, confectionary, candy store",
|
| 541 |
+
"n03095699": "container ship, containership, container vessel",
|
| 542 |
+
"n03100240": "convertible",
|
| 543 |
+
"n03109150": "corkscrew, bottle screw",
|
| 544 |
+
"n03110669": "cornet, horn, trumpet, trump",
|
| 545 |
+
"n03124043": "cowboy boot",
|
| 546 |
+
"n03124170": "cowboy hat, ten-gallon hat",
|
| 547 |
+
"n03125729": "cradle",
|
| 548 |
+
"n03126707": "crane2",
|
| 549 |
+
"n03127747": "crash helmet",
|
| 550 |
+
"n03127925": "crate",
|
| 551 |
+
"n03131574": "crib, cot",
|
| 552 |
+
"n03133878": "Crock Pot",
|
| 553 |
+
"n03134739": "croquet ball",
|
| 554 |
+
"n03141823": "crutch",
|
| 555 |
+
"n03146219": "cuirass",
|
| 556 |
+
"n03160309": "dam, dike, dyke",
|
| 557 |
+
"n03179701": "desk",
|
| 558 |
+
"n03180011": "desktop computer",
|
| 559 |
+
"n03187595": "dial telephone, dial phone",
|
| 560 |
+
"n03188531": "diaper, nappy, napkin",
|
| 561 |
+
"n03196217": "digital clock",
|
| 562 |
+
"n03197337": "digital watch",
|
| 563 |
+
"n03201208": "dining table, board",
|
| 564 |
+
"n03207743": "dishrag, dishcloth",
|
| 565 |
+
"n03207941": "dishwasher, dish washer, dishwashing machine",
|
| 566 |
+
"n03208938": "disk brake, disc brake",
|
| 567 |
+
"n03216828": "dock, dockage, docking facility",
|
| 568 |
+
"n03218198": "dogsled, dog sled, dog sleigh",
|
| 569 |
+
"n03220513": "dome",
|
| 570 |
+
"n03223299": "doormat, welcome mat",
|
| 571 |
+
"n03240683": "drilling platform, offshore rig",
|
| 572 |
+
"n03249569": "drum, membranophone, tympan",
|
| 573 |
+
"n03250847": "drumstick",
|
| 574 |
+
"n03255030": "dumbbell",
|
| 575 |
+
"n03259280": "Dutch oven",
|
| 576 |
+
"n03271574": "electric fan, blower",
|
| 577 |
+
"n03272010": "electric guitar",
|
| 578 |
+
"n03272562": "electric locomotive",
|
| 579 |
+
"n03290653": "entertainment center",
|
| 580 |
+
"n03291819": "envelope",
|
| 581 |
+
"n03297495": "espresso maker",
|
| 582 |
+
"n03314780": "face powder",
|
| 583 |
+
"n03325584": "feather boa, boa",
|
| 584 |
+
"n03337140": "file, file cabinet, filing cabinet",
|
| 585 |
+
"n03344393": "fireboat",
|
| 586 |
+
"n03345487": "fire engine, fire truck",
|
| 587 |
+
"n03347037": "fire screen, fireguard",
|
| 588 |
+
"n03355925": "flagpole, flagstaff",
|
| 589 |
+
"n03372029": "flute, transverse flute",
|
| 590 |
+
"n03376595": "folding chair",
|
| 591 |
+
"n03379051": "football helmet",
|
| 592 |
+
"n03384352": "forklift",
|
| 593 |
+
"n03388043": "fountain",
|
| 594 |
+
"n03388183": "fountain pen",
|
| 595 |
+
"n03388549": "four-poster",
|
| 596 |
+
"n03393912": "freight car",
|
| 597 |
+
"n03394916": "French horn, horn",
|
| 598 |
+
"n03400231": "frying pan, frypan, skillet",
|
| 599 |
+
"n03404251": "fur coat",
|
| 600 |
+
"n03417042": "garbage truck, dustcart",
|
| 601 |
+
"n03424325": "gasmask, respirator, gas helmet",
|
| 602 |
+
"n03425413": "gas pump, gasoline pump, petrol pump, island dispenser",
|
| 603 |
+
"n03443371": "goblet",
|
| 604 |
+
"n03444034": "go-kart",
|
| 605 |
+
"n03445777": "golf ball",
|
| 606 |
+
"n03445924": "golfcart, golf cart",
|
| 607 |
+
"n03447447": "gondola",
|
| 608 |
+
"n03447721": "gong, tam-tam",
|
| 609 |
+
"n03450230": "gown",
|
| 610 |
+
"n03452741": "grand piano, grand",
|
| 611 |
+
"n03457902": "greenhouse, nursery, glasshouse",
|
| 612 |
+
"n03459775": "grille, radiator grille",
|
| 613 |
+
"n03461385": "grocery store, grocery, food market, market",
|
| 614 |
+
"n03467068": "guillotine",
|
| 615 |
+
"n03476684": "hair slide",
|
| 616 |
+
"n03476991": "hair spray",
|
| 617 |
+
"n03478589": "half track",
|
| 618 |
+
"n03481172": "hammer",
|
| 619 |
+
"n03482405": "hamper",
|
| 620 |
+
"n03483316": "hand blower, blow dryer, blow drier, hair dryer, hair drier",
|
| 621 |
+
"n03485407": "hand-held computer, hand-held microcomputer",
|
| 622 |
+
"n03485794": "handkerchief, hankie, hanky, hankey",
|
| 623 |
+
"n03492542": "hard disc, hard disk, fixed disk",
|
| 624 |
+
"n03494278": "harmonica, mouth organ, harp, mouth harp",
|
| 625 |
+
"n03495258": "harp",
|
| 626 |
+
"n03496892": "harvester, reaper",
|
| 627 |
+
"n03498962": "hatchet",
|
| 628 |
+
"n03527444": "holster",
|
| 629 |
+
"n03529860": "home theater, home theatre",
|
| 630 |
+
"n03530642": "honeycomb",
|
| 631 |
+
"n03532672": "hook, claw",
|
| 632 |
+
"n03534580": "hoopskirt, crinoline",
|
| 633 |
+
"n03535780": "horizontal bar, high bar",
|
| 634 |
+
"n03538406": "horse cart, horse-cart",
|
| 635 |
+
"n03544143": "hourglass",
|
| 636 |
+
"n03584254": "iPod",
|
| 637 |
+
"n03584829": "iron, smoothing iron",
|
| 638 |
+
"n03590841": "jack-o'-lantern",
|
| 639 |
+
"n03594734": "jean, blue jean, denim",
|
| 640 |
+
"n03594945": "jeep, landrover",
|
| 641 |
+
"n03595614": "jersey, T-shirt, tee shirt",
|
| 642 |
+
"n03598930": "jigsaw puzzle",
|
| 643 |
+
"n03599486": "jinrikisha, ricksha, rickshaw",
|
| 644 |
+
"n03602883": "joystick",
|
| 645 |
+
"n03617480": "kimono",
|
| 646 |
+
"n03623198": "knee pad",
|
| 647 |
+
"n03627232": "knot",
|
| 648 |
+
"n03630383": "lab coat, laboratory coat",
|
| 649 |
+
"n03633091": "ladle",
|
| 650 |
+
"n03637318": "lampshade, lamp shade",
|
| 651 |
+
"n03642806": "laptop, laptop computer",
|
| 652 |
+
"n03649909": "lawn mower, mower",
|
| 653 |
+
"n03657121": "lens cap, lens cover",
|
| 654 |
+
"n03658185": "letter opener, paper knife, paperknife",
|
| 655 |
+
"n03661043": "library",
|
| 656 |
+
"n03662601": "lifeboat",
|
| 657 |
+
"n03666591": "lighter, light, igniter, ignitor",
|
| 658 |
+
"n03670208": "limousine, limo",
|
| 659 |
+
"n03673027": "liner, ocean liner",
|
| 660 |
+
"n03676483": "lipstick, lip rouge",
|
| 661 |
+
"n03680355": "Loafer",
|
| 662 |
+
"n03690938": "lotion",
|
| 663 |
+
"n03691459": "loudspeaker, speaker, speaker unit, loudspeaker system, speaker system",
|
| 664 |
+
"n03692522": "loupe, jeweler's loupe",
|
| 665 |
+
"n03697007": "lumbermill, sawmill",
|
| 666 |
+
"n03706229": "magnetic compass",
|
| 667 |
+
"n03709823": "mailbag, postbag",
|
| 668 |
+
"n03710193": "mailbox, letter box",
|
| 669 |
+
"n03710637": "maillot",
|
| 670 |
+
"n03710721": "maillot, tank suit",
|
| 671 |
+
"n03717622": "manhole cover",
|
| 672 |
+
"n03720891": "maraca",
|
| 673 |
+
"n03721384": "marimba, xylophone",
|
| 674 |
+
"n03724870": "mask",
|
| 675 |
+
"n03729826": "matchstick",
|
| 676 |
+
"n03733131": "maypole",
|
| 677 |
+
"n03733281": "maze, labyrinth",
|
| 678 |
+
"n03733805": "measuring cup",
|
| 679 |
+
"n03742115": "medicine chest, medicine cabinet",
|
| 680 |
+
"n03743016": "megalith, megalithic structure",
|
| 681 |
+
"n03759954": "microphone, mike",
|
| 682 |
+
"n03761084": "microwave, microwave oven",
|
| 683 |
+
"n03763968": "military uniform",
|
| 684 |
+
"n03764736": "milk can",
|
| 685 |
+
"n03769881": "minibus",
|
| 686 |
+
"n03770439": "miniskirt, mini",
|
| 687 |
+
"n03770679": "minivan",
|
| 688 |
+
"n03773504": "missile",
|
| 689 |
+
"n03775071": "mitten",
|
| 690 |
+
"n03775546": "mixing bowl",
|
| 691 |
+
"n03776460": "mobile home, manufactured home",
|
| 692 |
+
"n03777568": "Model T",
|
| 693 |
+
"n03777754": "modem",
|
| 694 |
+
"n03781244": "monastery",
|
| 695 |
+
"n03782006": "monitor",
|
| 696 |
+
"n03785016": "moped",
|
| 697 |
+
"n03786901": "mortar",
|
| 698 |
+
"n03787032": "mortarboard",
|
| 699 |
+
"n03788195": "mosque",
|
| 700 |
+
"n03788365": "mosquito net",
|
| 701 |
+
"n03791053": "motor scooter, scooter",
|
| 702 |
+
"n03792782": "mountain bike, all-terrain bike, off-roader",
|
| 703 |
+
"n03792972": "mountain tent",
|
| 704 |
+
"n03793489": "mouse, computer mouse",
|
| 705 |
+
"n03794056": "mousetrap",
|
| 706 |
+
"n03796401": "moving van",
|
| 707 |
+
"n03803284": "muzzle",
|
| 708 |
+
"n03804744": "nail",
|
| 709 |
+
"n03814639": "neck brace",
|
| 710 |
+
"n03814906": "necklace",
|
| 711 |
+
"n03825788": "nipple",
|
| 712 |
+
"n03832673": "notebook, notebook computer",
|
| 713 |
+
"n03837869": "obelisk",
|
| 714 |
+
"n03838899": "oboe, hautboy, hautbois",
|
| 715 |
+
"n03840681": "ocarina, sweet potato",
|
| 716 |
+
"n03841143": "odometer, hodometer, mileometer, milometer",
|
| 717 |
+
"n03843555": "oil filter",
|
| 718 |
+
"n03854065": "organ, pipe organ",
|
| 719 |
+
"n03857828": "oscilloscope, scope, cathode-ray oscilloscope, CRO",
|
| 720 |
+
"n03866082": "overskirt",
|
| 721 |
+
"n03868242": "oxcart",
|
| 722 |
+
"n03868863": "oxygen mask",
|
| 723 |
+
"n03871628": "packet",
|
| 724 |
+
"n03873416": "paddle, boat paddle",
|
| 725 |
+
"n03874293": "paddlewheel, paddle wheel",
|
| 726 |
+
"n03874599": "padlock",
|
| 727 |
+
"n03876231": "paintbrush",
|
| 728 |
+
"n03877472": "pajama, pyjama, pj's, jammies",
|
| 729 |
+
"n03877845": "palace",
|
| 730 |
+
"n03884397": "panpipe, pandean pipe, syrinx",
|
| 731 |
+
"n03887697": "paper towel",
|
| 732 |
+
"n03888257": "parachute, chute",
|
| 733 |
+
"n03888605": "parallel bars, bars",
|
| 734 |
+
"n03891251": "park bench",
|
| 735 |
+
"n03891332": "parking meter",
|
| 736 |
+
"n03895866": "passenger car, coach, carriage",
|
| 737 |
+
"n03899768": "patio, terrace",
|
| 738 |
+
"n03902125": "pay-phone, pay-station",
|
| 739 |
+
"n03903868": "pedestal, plinth, footstall",
|
| 740 |
+
"n03908618": "pencil box, pencil case",
|
| 741 |
+
"n03908714": "pencil sharpener",
|
| 742 |
+
"n03916031": "perfume, essence",
|
| 743 |
+
"n03920288": "Petri dish",
|
| 744 |
+
"n03924679": "photocopier",
|
| 745 |
+
"n03929660": "pick, plectrum, plectron",
|
| 746 |
+
"n03929855": "pickelhaube",
|
| 747 |
+
"n03930313": "picket fence, paling",
|
| 748 |
+
"n03930630": "pickup, pickup truck",
|
| 749 |
+
"n03933933": "pier",
|
| 750 |
+
"n03935335": "piggy bank, penny bank",
|
| 751 |
+
"n03937543": "pill bottle",
|
| 752 |
+
"n03938244": "pillow",
|
| 753 |
+
"n03942813": "ping-pong ball",
|
| 754 |
+
"n03944341": "pinwheel",
|
| 755 |
+
"n03947888": "pirate, pirate ship",
|
| 756 |
+
"n03950228": "pitcher, ewer",
|
| 757 |
+
"n03954731": "plane, carpenter's plane, woodworking plane",
|
| 758 |
+
"n03956157": "planetarium",
|
| 759 |
+
"n03958227": "plastic bag",
|
| 760 |
+
"n03961711": "plate rack",
|
| 761 |
+
"n03967562": "plow, plough",
|
| 762 |
+
"n03970156": "plunger, plumber's helper",
|
| 763 |
+
"n03976467": "Polaroid camera, Polaroid Land camera",
|
| 764 |
+
"n03976657": "pole",
|
| 765 |
+
"n03977966": "police van, police wagon, paddy wagon, patrol wagon, wagon, black Maria",
|
| 766 |
+
"n03980874": "poncho",
|
| 767 |
+
"n03982430": "pool table, billiard table, snooker table",
|
| 768 |
+
"n03983396": "pop bottle, soda bottle",
|
| 769 |
+
"n03991062": "pot, flowerpot",
|
| 770 |
+
"n03992509": "potter's wheel",
|
| 771 |
+
"n03995372": "power drill",
|
| 772 |
+
"n03998194": "prayer rug, prayer mat",
|
| 773 |
+
"n04004767": "printer",
|
| 774 |
+
"n04005630": "prison, prison house",
|
| 775 |
+
"n04008634": "projectile, missile",
|
| 776 |
+
"n04009552": "projector",
|
| 777 |
+
"n04019541": "puck, hockey puck",
|
| 778 |
+
"n04023962": "punching bag, punch bag, punching ball, punchball",
|
| 779 |
+
"n04026417": "purse",
|
| 780 |
+
"n04033901": "quill, quill pen",
|
| 781 |
+
"n04033995": "quilt, comforter, comfort, puff",
|
| 782 |
+
"n04037443": "racer, race car, racing car",
|
| 783 |
+
"n04039381": "racket, racquet",
|
| 784 |
+
"n04040759": "radiator",
|
| 785 |
+
"n04041544": "radio, wireless",
|
| 786 |
+
"n04044716": "radio telescope, radio reflector",
|
| 787 |
+
"n04049303": "rain barrel",
|
| 788 |
+
"n04065272": "recreational vehicle, RV, R.V.",
|
| 789 |
+
"n04067472": "reel",
|
| 790 |
+
"n04069434": "reflex camera",
|
| 791 |
+
"n04070727": "refrigerator, icebox",
|
| 792 |
+
"n04074963": "remote control, remote",
|
| 793 |
+
"n04081281": "restaurant, eating house, eating place, eatery",
|
| 794 |
+
"n04086273": "revolver, six-gun, six-shooter",
|
| 795 |
+
"n04090263": "rifle",
|
| 796 |
+
"n04099969": "rocking chair, rocker",
|
| 797 |
+
"n04111531": "rotisserie",
|
| 798 |
+
"n04116512": "rubber eraser, rubber, pencil eraser",
|
| 799 |
+
"n04118538": "rugby ball",
|
| 800 |
+
"n04118776": "rule, ruler",
|
| 801 |
+
"n04120489": "running shoe",
|
| 802 |
+
"n04125021": "safe",
|
| 803 |
+
"n04127249": "safety pin",
|
| 804 |
+
"n04131690": "saltshaker, salt shaker",
|
| 805 |
+
"n04133789": "sandal",
|
| 806 |
+
"n04136333": "sarong",
|
| 807 |
+
"n04141076": "sax, saxophone",
|
| 808 |
+
"n04141327": "scabbard",
|
| 809 |
+
"n04141975": "scale, weighing machine",
|
| 810 |
+
"n04146614": "school bus",
|
| 811 |
+
"n04147183": "schooner",
|
| 812 |
+
"n04149813": "scoreboard",
|
| 813 |
+
"n04152593": "screen, CRT screen",
|
| 814 |
+
"n04153751": "screw",
|
| 815 |
+
"n04154565": "screwdriver",
|
| 816 |
+
"n04162706": "seat belt, seatbelt",
|
| 817 |
+
"n04179913": "sewing machine",
|
| 818 |
+
"n04192698": "shield, buckler",
|
| 819 |
+
"n04200800": "shoe shop, shoe-shop, shoe store",
|
| 820 |
+
"n04201297": "shoji",
|
| 821 |
+
"n04204238": "shopping basket",
|
| 822 |
+
"n04204347": "shopping cart",
|
| 823 |
+
"n04208210": "shovel",
|
| 824 |
+
"n04209133": "shower cap",
|
| 825 |
+
"n04209239": "shower curtain",
|
| 826 |
+
"n04228054": "ski",
|
| 827 |
+
"n04229816": "ski mask",
|
| 828 |
+
"n04235860": "sleeping bag",
|
| 829 |
+
"n04238763": "slide rule, slipstick",
|
| 830 |
+
"n04239074": "sliding door",
|
| 831 |
+
"n04243546": "slot, one-armed bandit",
|
| 832 |
+
"n04251144": "snorkel",
|
| 833 |
+
"n04252077": "snowmobile",
|
| 834 |
+
"n04252225": "snowplow, snowplough",
|
| 835 |
+
"n04254120": "soap dispenser",
|
| 836 |
+
"n04254680": "soccer ball",
|
| 837 |
+
"n04254777": "sock",
|
| 838 |
+
"n04258138": "solar dish, solar collector, solar furnace",
|
| 839 |
+
"n04259630": "sombrero",
|
| 840 |
+
"n04263257": "soup bowl",
|
| 841 |
+
"n04264628": "space bar",
|
| 842 |
+
"n04265275": "space heater",
|
| 843 |
+
"n04266014": "space shuttle",
|
| 844 |
+
"n04270147": "spatula",
|
| 845 |
+
"n04273569": "speedboat",
|
| 846 |
+
"n04275548": "spider web, spider's web",
|
| 847 |
+
"n04277352": "spindle",
|
| 848 |
+
"n04285008": "sports car, sport car",
|
| 849 |
+
"n04286575": "spotlight, spot",
|
| 850 |
+
"n04296562": "stage",
|
| 851 |
+
"n04310018": "steam locomotive",
|
| 852 |
+
"n04311004": "steel arch bridge",
|
| 853 |
+
"n04311174": "steel drum",
|
| 854 |
+
"n04317175": "stethoscope",
|
| 855 |
+
"n04325704": "stole",
|
| 856 |
+
"n04326547": "stone wall",
|
| 857 |
+
"n04328186": "stopwatch, stop watch",
|
| 858 |
+
"n04330267": "stove",
|
| 859 |
+
"n04332243": "strainer",
|
| 860 |
+
"n04335435": "streetcar, tram, tramcar, trolley, trolley car",
|
| 861 |
+
"n04336792": "stretcher",
|
| 862 |
+
"n04344873": "studio couch, day bed",
|
| 863 |
+
"n04346328": "stupa, tope",
|
| 864 |
+
"n04347754": "submarine, pigboat, sub, U-boat",
|
| 865 |
+
"n04350905": "suit, suit of clothes",
|
| 866 |
+
"n04355338": "sundial",
|
| 867 |
+
"n04355933": "sunglass",
|
| 868 |
+
"n04356056": "sunglasses, dark glasses, shades",
|
| 869 |
+
"n04357314": "sunscreen, sunblock, sun blocker",
|
| 870 |
+
"n04366367": "suspension bridge",
|
| 871 |
+
"n04367480": "swab, swob, mop",
|
| 872 |
+
"n04370456": "sweatshirt",
|
| 873 |
+
"n04371430": "swimming trunks, bathing trunks",
|
| 874 |
+
"n04371774": "swing",
|
| 875 |
+
"n04372370": "switch, electric switch, electrical switch",
|
| 876 |
+
"n04376876": "syringe",
|
| 877 |
+
"n04380533": "table lamp",
|
| 878 |
+
"n04389033": "tank, army tank, armored combat vehicle, armoured combat vehicle",
|
| 879 |
+
"n04392985": "tape player",
|
| 880 |
+
"n04398044": "teapot",
|
| 881 |
+
"n04399382": "teddy, teddy bear",
|
| 882 |
+
"n04404412": "television, television system",
|
| 883 |
+
"n04409515": "tennis ball",
|
| 884 |
+
"n04417672": "thatch, thatched roof",
|
| 885 |
+
"n04418357": "theater curtain, theatre curtain",
|
| 886 |
+
"n04423845": "thimble",
|
| 887 |
+
"n04428191": "thresher, thrasher, threshing machine",
|
| 888 |
+
"n04429376": "throne",
|
| 889 |
+
"n04435653": "tile roof",
|
| 890 |
+
"n04442312": "toaster",
|
| 891 |
+
"n04443257": "tobacco shop, tobacconist shop, tobacconist",
|
| 892 |
+
"n04447861": "toilet seat",
|
| 893 |
+
"n04456115": "torch",
|
| 894 |
+
"n04458633": "totem pole",
|
| 895 |
+
"n04461696": "tow truck, tow car, wrecker",
|
| 896 |
+
"n04462240": "toyshop",
|
| 897 |
+
"n04465501": "tractor",
|
| 898 |
+
"n04467665": "trailer truck, tractor trailer, trucking rig, rig, articulated lorry, semi",
|
| 899 |
+
"n04476259": "tray",
|
| 900 |
+
"n04479046": "trench coat",
|
| 901 |
+
"n04482393": "tricycle, trike, velocipede",
|
| 902 |
+
"n04483307": "trimaran",
|
| 903 |
+
"n04485082": "tripod",
|
| 904 |
+
"n04486054": "triumphal arch",
|
| 905 |
+
"n04487081": "trolleybus, trolley coach, trackless trolley",
|
| 906 |
+
"n04487394": "trombone",
|
| 907 |
+
"n04493381": "tub, vat",
|
| 908 |
+
"n04501370": "turnstile",
|
| 909 |
+
"n04505470": "typewriter keyboard",
|
| 910 |
+
"n04507155": "umbrella",
|
| 911 |
+
"n04509417": "unicycle, monocycle",
|
| 912 |
+
"n04515003": "upright, upright piano",
|
| 913 |
+
"n04517823": "vacuum, vacuum cleaner",
|
| 914 |
+
"n04522168": "vase",
|
| 915 |
+
"n04523525": "vault",
|
| 916 |
+
"n04525038": "velvet",
|
| 917 |
+
"n04525305": "vending machine",
|
| 918 |
+
"n04532106": "vestment",
|
| 919 |
+
"n04532670": "viaduct",
|
| 920 |
+
"n04536866": "violin, fiddle",
|
| 921 |
+
"n04540053": "volleyball",
|
| 922 |
+
"n04542943": "waffle iron",
|
| 923 |
+
"n04548280": "wall clock",
|
| 924 |
+
"n04548362": "wallet, billfold, notecase, pocketbook",
|
| 925 |
+
"n04550184": "wardrobe, closet, press",
|
| 926 |
+
"n04552348": "warplane, military plane",
|
| 927 |
+
"n04553703": "washbasin, handbasin, washbowl, lavabo, wash-hand basin",
|
| 928 |
+
"n04554684": "washer, automatic washer, washing machine",
|
| 929 |
+
"n04557648": "water bottle",
|
| 930 |
+
"n04560804": "water jug",
|
| 931 |
+
"n04562935": "water tower",
|
| 932 |
+
"n04579145": "whiskey jug",
|
| 933 |
+
"n04579432": "whistle",
|
| 934 |
+
"n04584207": "wig",
|
| 935 |
+
"n04589890": "window screen",
|
| 936 |
+
"n04590129": "window shade",
|
| 937 |
+
"n04591157": "Windsor tie",
|
| 938 |
+
"n04591713": "wine bottle",
|
| 939 |
+
"n04592741": "wing",
|
| 940 |
+
"n04596742": "wok",
|
| 941 |
+
"n04597913": "wooden spoon",
|
| 942 |
+
"n04599235": "wool, woolen, woollen",
|
| 943 |
+
"n04604644": "worm fence, snake fence, snake-rail fence, Virginia fence",
|
| 944 |
+
"n04606251": "wreck",
|
| 945 |
+
"n04612504": "yawl",
|
| 946 |
+
"n04613696": "yurt",
|
| 947 |
+
"n06359193": "web site, website, internet site, site",
|
| 948 |
+
"n06596364": "comic book",
|
| 949 |
+
"n06785654": "crossword puzzle, crossword",
|
| 950 |
+
"n06794110": "street sign",
|
| 951 |
+
"n06874185": "traffic light, traffic signal, stoplight",
|
| 952 |
+
"n07248320": "book jacket, dust cover, dust jacket, dust wrapper",
|
| 953 |
+
"n07565083": "menu",
|
| 954 |
+
"n07579787": "plate",
|
| 955 |
+
"n07583066": "guacamole",
|
| 956 |
+
"n07584110": "consomme",
|
| 957 |
+
"n07590611": "hot pot, hotpot",
|
| 958 |
+
"n07613480": "trifle",
|
| 959 |
+
"n07614500": "ice cream, icecream",
|
| 960 |
+
"n07615774": "ice lolly, lolly, lollipop, popsicle",
|
| 961 |
+
"n07684084": "French loaf",
|
| 962 |
+
"n07693725": "bagel, beigel",
|
| 963 |
+
"n07695742": "pretzel",
|
| 964 |
+
"n07697313": "cheeseburger",
|
| 965 |
+
"n07697537": "hotdog, hot dog, red hot",
|
| 966 |
+
"n07711569": "mashed potato",
|
| 967 |
+
"n07714571": "head cabbage",
|
| 968 |
+
"n07714990": "broccoli",
|
| 969 |
+
"n07715103": "cauliflower",
|
| 970 |
+
"n07716358": "zucchini, courgette",
|
| 971 |
+
"n07716906": "spaghetti squash",
|
| 972 |
+
"n07717410": "acorn squash",
|
| 973 |
+
"n07717556": "butternut squash",
|
| 974 |
+
"n07718472": "cucumber, cuke",
|
| 975 |
+
"n07718747": "artichoke, globe artichoke",
|
| 976 |
+
"n07720875": "bell pepper",
|
| 977 |
+
"n07730033": "cardoon",
|
| 978 |
+
"n07734744": "mushroom",
|
| 979 |
+
"n07742313": "Granny Smith",
|
| 980 |
+
"n07745940": "strawberry",
|
| 981 |
+
"n07747607": "orange",
|
| 982 |
+
"n07749582": "lemon",
|
| 983 |
+
"n07753113": "fig",
|
| 984 |
+
"n07753275": "pineapple, ananas",
|
| 985 |
+
"n07753592": "banana",
|
| 986 |
+
"n07754684": "jackfruit, jak, jack",
|
| 987 |
+
"n07760859": "custard apple",
|
| 988 |
+
"n07768694": "pomegranate",
|
| 989 |
+
"n07802026": "hay",
|
| 990 |
+
"n07831146": "carbonara",
|
| 991 |
+
"n07836838": "chocolate sauce, chocolate syrup",
|
| 992 |
+
"n07860988": "dough",
|
| 993 |
+
"n07871810": "meat loaf, meatloaf",
|
| 994 |
+
"n07873807": "pizza, pizza pie",
|
| 995 |
+
"n07875152": "potpie",
|
| 996 |
+
"n07880968": "burrito",
|
| 997 |
+
"n07892512": "red wine",
|
| 998 |
+
"n07920052": "espresso",
|
| 999 |
+
"n07930864": "cup",
|
| 1000 |
+
"n07932039": "eggnog",
|
| 1001 |
+
"n09193705": "alp",
|
| 1002 |
+
"n09229709": "bubble",
|
| 1003 |
+
"n09246464": "cliff, drop, drop-off",
|
| 1004 |
+
"n09256479": "coral reef",
|
| 1005 |
+
"n09288635": "geyser",
|
| 1006 |
+
"n09332890": "lakeside, lakeshore",
|
| 1007 |
+
"n09399592": "promontory, headland, head, foreland",
|
| 1008 |
+
"n09421951": "sandbar, sand bar",
|
| 1009 |
+
"n09428293": "seashore, coast, seacoast, sea-coast",
|
| 1010 |
+
"n09468604": "valley, vale",
|
| 1011 |
+
"n09472597": "volcano",
|
| 1012 |
+
"n09835506": "ballplayer, baseball player",
|
| 1013 |
+
"n10148035": "groom, bridegroom",
|
| 1014 |
+
"n10565667": "scuba diver",
|
| 1015 |
+
"n11879895": "rapeseed",
|
| 1016 |
+
"n11939491": "daisy",
|
| 1017 |
+
"n12057211": "yellow lady's slipper, yellow lady-slipper, Cypripedium calceolus, Cypripedium parviflorum",
|
| 1018 |
+
"n12144580": "corn",
|
| 1019 |
+
"n12267677": "acorn",
|
| 1020 |
+
"n12620546": "hip, rose hip, rosehip",
|
| 1021 |
+
"n12768682": "buckeye, horse chestnut, conker",
|
| 1022 |
+
"n12985857": "coral fungus",
|
| 1023 |
+
"n12998815": "agaric",
|
| 1024 |
+
"n13037406": "gyromitra",
|
| 1025 |
+
"n13040303": "stinkhorn, carrion fungus",
|
| 1026 |
+
"n13044778": "earthstar",
|
| 1027 |
+
"n13052670": "hen-of-the-woods, hen of the woods, Polyporus frondosus, Grifola frondosa",
|
| 1028 |
+
"n13054560": "bolete",
|
| 1029 |
+
"n13133613": "ear, spike, capitulum",
|
| 1030 |
+
"n15075141": "toilet tissue, toilet paper, bathroom tissue",
|
| 1031 |
+
}
|
| 1032 |
+
)
|
| 1033 |
+
|
| 1034 |
+
_CITATION = """\
|
| 1035 |
+
@misc{nauen2025foraug,
|
| 1036 |
+
title={ForAug: Recombining Foregrounds and Backgrounds to Improve Vision Transformer Training with Bias Mitigation},
|
| 1037 |
+
author={Tobias Christian Nauen and Brian Moser and Federico Raue and Stanislav Frolov and Andreas Dengel},
|
| 1038 |
+
year={2025},
|
| 1039 |
+
eprint={2503.09399},
|
| 1040 |
+
archivePrefix={arXiv},
|
| 1041 |
+
primaryClass={cs.CV},
|
| 1042 |
+
}
|
| 1043 |
+
"""
|
| 1044 |
+
|
| 1045 |
+
_DESCRIPTION = """\
|
| 1046 |
+
ForNet is a dataset of foreground objects and backgrounds extracted (and infilled) from ImageNet. \
|
| 1047 |
+
It's the output of the segmentation phase of the ForAug data augmentation. \
|
| 1048 |
+
ForNet recombines these foregrounds and backgrounds on the fly to create new samples for training vision transformers.
|
| 1049 |
+
"""
|
| 1050 |
+
|
| 1051 |
+
_GIT = "https://github.com/tobna/ForAug"
|
| 1052 |
+
_HOMEPAGE = "Coming Soon"
|
| 1053 |
+
_DATASET_URL = "https://huggingface.co/datasets/TNauen/ForNet/resolve/main/"
|
| 1054 |
+
_CONST_URLS = (
|
| 1055 |
+
[_DATASET_URL + "settings.txt"]
|
| 1056 |
+
+ [_DATASET_URL + f"fg_bg_ratios_{part}.json" for part in ["train", "val"]]
|
| 1057 |
+
+ [_DATASET_URL + f"hf_{part}_indices.json" for part in ["train", "val"]]
|
| 1058 |
+
)
|
| 1059 |
+
_PATCH_URLS = [_DATASET_URL + f"train_{i}.zip" for i in range(20)] + [_DATASET_URL + "val.zip"]
|
| 1060 |
+
|
| 1061 |
+
|
| 1062 |
+
class RecombineDataset(Dataset):
|
| 1063 |
+
"""Wrapper for ForNet dataset that recombines foregrounds and backgrounds on the fly."""
|
| 1064 |
+
|
| 1065 |
+
def __init__(
|
| 1066 |
+
self,
|
| 1067 |
+
*args,
|
| 1068 |
+
background_combination,
|
| 1069 |
+
fg_scale_jitter,
|
| 1070 |
+
pruning_ratio,
|
| 1071 |
+
fg_size_mode,
|
| 1072 |
+
fg_bates_n,
|
| 1073 |
+
mask_smoothing_sigma,
|
| 1074 |
+
rel_jut_out,
|
| 1075 |
+
orig_img_prob,
|
| 1076 |
+
**kwargs,
|
| 1077 |
+
):
|
| 1078 |
+
"""Create the ForNet recombination dataset.
|
| 1079 |
+
|
| 1080 |
+
Args:
|
| 1081 |
+
background_combination (str): Which backgrounds to combine with foregrounds. Options: "orig", "same", "all".
|
| 1082 |
+
fg_scale_jitter (tuple[float]): How much should the size of the foreground be changed (random ratio). Example: (0.1, 0.8).
|
| 1083 |
+
pruning_ratio (float): For pruning backgrounds, with (foreground size/background size) >= <pruning_ratio>. Backgrounds from images that contain very large foreground objects are mostly computer generated and therefore relatively unnatural. Full dataset: 1.1 .
|
| 1084 |
+
fg_size_mode (str): How to determine the size of the foreground, based on the foreground sizes of the foreground and background images. Options: "range", "min", "max", "mean".
|
| 1085 |
+
fg_bates_n (int): Bates parameter for the distribution of the object position in the foreground. Uniform Distribution: 1. The higher the value, the more likely the object is in the center. For fg_bates_n = 0, the object is always in the center.
|
| 1086 |
+
mask_smoothing_sigma (float): Sigma for the Gaussian blur of the mask edge.
|
| 1087 |
+
rel_jut_out (float): How much is the foreground allowed to stand/jut out of the background (and then cut off).
|
| 1088 |
+
orig_img_prob (float | str): Probability to use the original image, instead of the fg-bg recombinations. Options: 0.0-1.0, "linear", "revlinear", "cos".
|
| 1089 |
+
"""
|
| 1090 |
+
super().__init__(*args, **kwargs)
|
| 1091 |
+
assert (isinstance(orig_img_prob, float) and 0.0 <= orig_img_prob <= 1.0) or orig_img_prob in [
|
| 1092 |
+
"linear",
|
| 1093 |
+
"revlinear",
|
| 1094 |
+
"cos",
|
| 1095 |
+
], f"Invalid orig_img_prob {orig_img_prob}"
|
| 1096 |
+
assert background_combination in [
|
| 1097 |
+
"all",
|
| 1098 |
+
"same",
|
| 1099 |
+
"orig",
|
| 1100 |
+
], f"Invalid background_combination {background_combination}"
|
| 1101 |
+
assert fg_size_mode in ["range", "min", "max", "mean"], f"Invalid fg_size_mode {fg_size_mode}"
|
| 1102 |
+
self.background_combination = background_combination
|
| 1103 |
+
self.fg_scale_jitter = fg_scale_jitter
|
| 1104 |
+
self.pruning_ratio = pruning_ratio
|
| 1105 |
+
self.fg_size_mode = fg_size_mode
|
| 1106 |
+
self.fg_bates_n = fg_bates_n
|
| 1107 |
+
self.mask_smoothing_sigma = mask_smoothing_sigma
|
| 1108 |
+
self.rel_jut_out = rel_jut_out
|
| 1109 |
+
self.orig_img_prob = orig_img_prob
|
| 1110 |
+
self.epochs = 0
|
| 1111 |
+
self._epoch = 0
|
| 1112 |
+
self.cls_to_idx = {}
|
| 1113 |
+
|
| 1114 |
+
bg_rat_indices = super()._getitem(0)["bg_rat_idx_file"]
|
| 1115 |
+
self.train = "train" in bg_rat_indices.split("/")[-1]
|
| 1116 |
+
|
| 1117 |
+
with open(bg_rat_indices, "r") as f:
|
| 1118 |
+
bg_rat_indices = json.load(f)
|
| 1119 |
+
for in_cls in bg_rat_indices:
|
| 1120 |
+
if in_cls not in self.cls_to_idx:
|
| 1121 |
+
self.cls_to_idx[in_cls] = []
|
| 1122 |
+
for idx, rat in bg_rat_indices[in_cls]:
|
| 1123 |
+
if rat < self.pruning_ratio:
|
| 1124 |
+
self.cls_to_idx[in_cls].append(idx)
|
| 1125 |
+
if self.background_combination == "all":
|
| 1126 |
+
self.cls_to_idx["all"] = list(itertools.chain(*self.cls_to_idx.values()))
|
| 1127 |
+
|
| 1128 |
+
@property
|
| 1129 |
+
def total_epochs(self):
|
| 1130 |
+
return self.epochs
|
| 1131 |
+
|
| 1132 |
+
@total_epochs.setter
|
| 1133 |
+
def total_epochs(self, value):
|
| 1134 |
+
self.epochs = value
|
| 1135 |
+
|
| 1136 |
+
@property
|
| 1137 |
+
def epoch(self):
|
| 1138 |
+
return self._epoch
|
| 1139 |
+
|
| 1140 |
+
@epoch.setter
|
| 1141 |
+
def epoch(self, value):
|
| 1142 |
+
assert 0 <= value < self.epochs, f"Epoch {value} is out of bounds for range [0, {self.epochs})"
|
| 1143 |
+
self._epoch = value
|
| 1144 |
+
|
| 1145 |
+
def _getitem(self, key):
|
| 1146 |
+
fg_item = super()._getitem(key)
|
| 1147 |
+
out_dict = {"label": fg_item["label"]}
|
| 1148 |
+
in_cls = fg_item["path"].split("/")[0]
|
| 1149 |
+
|
| 1150 |
+
if (
|
| 1151 |
+
(self.orig_img_prob == "linear" and np.random.rand() < self._epoch / self.epochs)
|
| 1152 |
+
or (self.orig_img_prob == "revlinear" and np.random.rand() < (self._epoch - self.epochs) / self.epochs)
|
| 1153 |
+
or (self.orig_img_prob == "cos" and np.random.rand() > np.cos(np.pi * self._epoch / (2 * self.epochs)))
|
| 1154 |
+
or (
|
| 1155 |
+
isinstance(self.orig_img_prob, float)
|
| 1156 |
+
and self.orig_img_prob > 0.0
|
| 1157 |
+
and np.random.rand() < self.orig_img_prob
|
| 1158 |
+
)
|
| 1159 |
+
):
|
| 1160 |
+
# return original image
|
| 1161 |
+
out_dict["image"] = fg_item["in"]
|
| 1162 |
+
return out_dict
|
| 1163 |
+
|
| 1164 |
+
if self.background_combination == "orig":
|
| 1165 |
+
bg_item = fg_item
|
| 1166 |
+
elif self.background_combination == "same":
|
| 1167 |
+
rand_idx = np.random.randint(len(self.cls_to_idx[in_cls]))
|
| 1168 |
+
rand_idx = self.cls_to_idx[in_cls][rand_idx]
|
| 1169 |
+
bg_item = super()._getitem(rand_idx)
|
| 1170 |
+
else:
|
| 1171 |
+
# all
|
| 1172 |
+
rand_idx = np.random.randint(len(self.cls_to_idx["all"]))
|
| 1173 |
+
rand_idx = self.cls_to_idx["all"][rand_idx]
|
| 1174 |
+
bg_item = super()._getitem(rand_idx)
|
| 1175 |
+
|
| 1176 |
+
fg_img = fg_item["fg"].convert("RGBA")
|
| 1177 |
+
bg_img = bg_item["bg"].convert("RGB")
|
| 1178 |
+
bg_size = bg_img.size
|
| 1179 |
+
bg_area = bg_size[0] * bg_size[1]
|
| 1180 |
+
orig_fg_ratio = fg_item["fg/bg_area"]
|
| 1181 |
+
bg_fg_ratio = bg_item["fg/bg_area"]
|
| 1182 |
+
|
| 1183 |
+
if self.fg_size_mode == "max":
|
| 1184 |
+
goal_fg_ratio_lower = goal_fg_ratio_upper = max(orig_fg_ratio, bg_fg_ratio)
|
| 1185 |
+
elif self.fg_size_mode == "min":
|
| 1186 |
+
goal_fg_ratio_lower = goal_fg_ratio_upper = min(orig_fg_ratio, bg_fg_ratio)
|
| 1187 |
+
elif self.fg_size_mode == "mean":
|
| 1188 |
+
goal_fg_ratio_lower = goal_fg_ratio_upper = (orig_fg_ratio + bg_fg_ratio) / 2
|
| 1189 |
+
else:
|
| 1190 |
+
# range
|
| 1191 |
+
goal_fg_ratio_lower = min(orig_fg_ratio, bg_fg_ratio)
|
| 1192 |
+
goal_fg_ratio_upper = max(orig_fg_ratio, bg_fg_ratio)
|
| 1193 |
+
|
| 1194 |
+
fg_size_factor = T.ToTensor()(fg_img.split()[-1]).mean().item()
|
| 1195 |
+
|
| 1196 |
+
fg_scale = (
|
| 1197 |
+
np.random.uniform(
|
| 1198 |
+
goal_fg_ratio_lower * (1 - self.fg_scale_jitter), goal_fg_ratio_upper * (1 + self.fg_scale_jitter)
|
| 1199 |
+
)
|
| 1200 |
+
/ fg_size_factor
|
| 1201 |
+
)
|
| 1202 |
+
|
| 1203 |
+
goal_shape_y = round(np.sqrt(bg_area * fg_scale * fg_img.size[1] / fg_img.size[0]))
|
| 1204 |
+
goal_shape_x = round(np.sqrt(bg_area * fg_scale * fg_img.size[0] / fg_img.size[1]))
|
| 1205 |
+
|
| 1206 |
+
fg_img = fg_img.resize((goal_shape_x, goal_shape_y))
|
| 1207 |
+
|
| 1208 |
+
if fg_img.size[0] > bg_size[0] or fg_img.size[1] > bg_size[1]:
|
| 1209 |
+
# random crop to fit
|
| 1210 |
+
goal_w, goal_h = (min(fg_img.size[0], bg_size[0]), min(fg_img.size[1], bg_size[1]))
|
| 1211 |
+
fg_img = T.RandomCrop((goal_h, goal_w))(fg_img) if self.train else T.CenterCrop((goal_h, goal_w))(fg_img)
|
| 1212 |
+
|
| 1213 |
+
# paste fg on bg
|
| 1214 |
+
z1, z2 = (
|
| 1215 |
+
(
|
| 1216 |
+
np.random.uniform(0, 1, abs(self.fg_bates_n)).mean(), # bates distribution n=1 => uniform
|
| 1217 |
+
np.random.uniform(0, 1, abs(self.fg_bates_n)).mean(),
|
| 1218 |
+
)
|
| 1219 |
+
if self.fg_bates_n != 0
|
| 1220 |
+
else (0.5, 0.5)
|
| 1221 |
+
)
|
| 1222 |
+
if self.fg_bates_n < 0:
|
| 1223 |
+
z1 = z1 + 0.5 - floor(z1 + 0.5)
|
| 1224 |
+
z2 = z2 + 0.5 - floor(z2 + 0.5)
|
| 1225 |
+
|
| 1226 |
+
x_min = -self.rel_jut_out * fg_img.size[0]
|
| 1227 |
+
x_max = bg_size[0] - fg_img.size[0] * (1 - self.rel_jut_out)
|
| 1228 |
+
y_min = -self.rel_jut_out * fg_img.size[1]
|
| 1229 |
+
y_max = bg_size[1] - fg_img.size[1] * (1 - self.rel_jut_out)
|
| 1230 |
+
|
| 1231 |
+
if x_min > x_max:
|
| 1232 |
+
x_min = x_max = (x_min + x_max) / 2
|
| 1233 |
+
if y_min > y_max:
|
| 1234 |
+
y_min = y_max = (y_min + y_max) / 2
|
| 1235 |
+
|
| 1236 |
+
offs_x = round(z1 * (x_max - x_min) + x_min)
|
| 1237 |
+
offs_y = round(z2 * (y_max - y_min) + y_min)
|
| 1238 |
+
|
| 1239 |
+
paste_mask = fg_img.split()[-1]
|
| 1240 |
+
if self.mask_smoothing_sigma > 0.0:
|
| 1241 |
+
sigma = (np.random.rand() * 0.9 + 0.1) * self.mask_smoothing_sigma
|
| 1242 |
+
paste_mask = paste_mask.filter(ImageFilter.GaussianBlur(radius=sigma))
|
| 1243 |
+
paste_mask = paste_mask.point(lambda p: 2 * p - 255 if p > 128 else 0)
|
| 1244 |
+
|
| 1245 |
+
bg_img.paste(fg_img.convert("RGB"), (offs_x, offs_y), paste_mask)
|
| 1246 |
+
bg_img = bg_img.convert("RGB")
|
| 1247 |
+
|
| 1248 |
+
out_dict["image"] = bg_img
|
| 1249 |
+
|
| 1250 |
+
return out_dict
|
| 1251 |
+
|
| 1252 |
+
def __str__(self):
|
| 1253 |
+
return f"{self.__class__}(\n\t features: ['image', 'label'],\n\t num_rows: {len(self)},\n\tbackground_combination: {self.background_combination},\n\t pruning_ratio: {self.pruning_ratio},\n\t fg_size_mode: {self.fg_size_mode},\n\t mask_smoothing_sigma: {self.mask_smoothing_sigma},\n\t orig_img_prob: {self.orig_img_prob}\n)"
|
| 1254 |
+
|
| 1255 |
+
|
| 1256 |
+
_CONFIG_HASH_IGNORE_KWARGS = [
|
| 1257 |
+
"background_combination",
|
| 1258 |
+
"fg_scale_jitter",
|
| 1259 |
+
"pruning_ratio",
|
| 1260 |
+
"fg_size_mode",
|
| 1261 |
+
"fg_bates_n",
|
| 1262 |
+
"mask_smoothing_sigma",
|
| 1263 |
+
"rel_jut_out",
|
| 1264 |
+
"orig_img_prob",
|
| 1265 |
+
]
|
| 1266 |
+
|
| 1267 |
+
|
| 1268 |
+
class ForNetConfig(datasets.BuilderConfig):
|
| 1269 |
+
"""BuilderConfig for ForNet."""
|
| 1270 |
+
|
| 1271 |
+
def __init__(
|
| 1272 |
+
self,
|
| 1273 |
+
background_combination,
|
| 1274 |
+
fg_scale_jitter,
|
| 1275 |
+
pruning_ratio,
|
| 1276 |
+
fg_size_mode,
|
| 1277 |
+
fg_bates_n,
|
| 1278 |
+
mask_smoothing_sigma,
|
| 1279 |
+
rel_jut_out,
|
| 1280 |
+
orig_img_prob,
|
| 1281 |
+
**kwargs,
|
| 1282 |
+
):
|
| 1283 |
+
"""BuilderConfig for ForNet.
|
| 1284 |
+
|
| 1285 |
+
Args:
|
| 1286 |
+
**kwargs: keyword arguments forwarded to super.
|
| 1287 |
+
"""
|
| 1288 |
+
super(ForNetConfig, self).__init__(**kwargs)
|
| 1289 |
+
self.background_combination = background_combination
|
| 1290 |
+
self.fg_scale_jitter = fg_scale_jitter
|
| 1291 |
+
self.pruning_ratio = pruning_ratio
|
| 1292 |
+
self.fg_size_mode = fg_size_mode
|
| 1293 |
+
self.fg_bates_n = fg_bates_n
|
| 1294 |
+
self.mask_smoothing_sigma = mask_smoothing_sigma
|
| 1295 |
+
self.rel_jut_out = rel_jut_out
|
| 1296 |
+
self.orig_img_prob = orig_img_prob
|
| 1297 |
+
|
| 1298 |
+
def __str__(self):
|
| 1299 |
+
return f"ForNetConfig(name={self.name}, version={self.version}, data_dir={self.data_dir}, data_files={self.data_files}, description={self.description}, background_combination={self.background_combination}, fg_scale_jitter={self.fg_scale_jitter}, pruning_ratio={self.pruning_ratio}, fg_size_mode={self.fg_size_mode}, fg_bates_n={self.fg_bates_n}, mask_smoothing_sigma={self.mask_smoothing_sigma}, rel_jut_out={self.rel_jut_out}, orig_img_prob={self.orig_img_prob})"
|
| 1300 |
+
|
| 1301 |
+
def create_config_id(
|
| 1302 |
+
self,
|
| 1303 |
+
config_kwargs: dict,
|
| 1304 |
+
custom_features=None,
|
| 1305 |
+
) -> str:
|
| 1306 |
+
"""The config id is used to build the cache directory.
|
| 1307 |
+
|
| 1308 |
+
By default it is equal to the config name.
|
| 1309 |
+
However the name of a config is not sufficient to have a unique identifier for the dataset being generated
|
| 1310 |
+
since it doesn't take into account:
|
| 1311 |
+
- the config kwargs that can be used to overwrite attributes
|
| 1312 |
+
- the custom features used to write the dataset
|
| 1313 |
+
- the data_files for json/text/csv/pandas datasets.
|
| 1314 |
+
|
| 1315 |
+
Therefore the config id is just the config name with an optional suffix based on these.
|
| 1316 |
+
"""
|
| 1317 |
+
# Possibly add a suffix to the name to handle custom features/data_files/config_kwargs
|
| 1318 |
+
suffix: Optional[str] = None
|
| 1319 |
+
config_kwargs_to_add_to_suffix = config_kwargs.copy()
|
| 1320 |
+
# name and version are already used to build the cache directory
|
| 1321 |
+
config_kwargs_to_add_to_suffix.pop("name", None)
|
| 1322 |
+
config_kwargs_to_add_to_suffix.pop("version", None)
|
| 1323 |
+
|
| 1324 |
+
# remove only recombination-relevant values
|
| 1325 |
+
for k in _CONFIG_HASH_IGNORE_KWARGS:
|
| 1326 |
+
config_kwargs_to_add_to_suffix.pop(k, None)
|
| 1327 |
+
|
| 1328 |
+
# data dir handling (when specified it points to the manually downloaded data):
|
| 1329 |
+
# it was previously ignored before the introduction of config id because we didn't want
|
| 1330 |
+
# to change the config name. Now it's fine to take it into account for the config id.
|
| 1331 |
+
# config_kwargs_to_add_to_suffix.pop("data_dir", None)
|
| 1332 |
+
if "data_dir" in config_kwargs_to_add_to_suffix:
|
| 1333 |
+
if config_kwargs_to_add_to_suffix["data_dir"] is None:
|
| 1334 |
+
config_kwargs_to_add_to_suffix.pop("data_dir", None)
|
| 1335 |
+
else:
|
| 1336 |
+
# canonicalize the data dir to avoid two paths to the same location having different
|
| 1337 |
+
# hashes
|
| 1338 |
+
data_dir = config_kwargs_to_add_to_suffix["data_dir"]
|
| 1339 |
+
data_dir = os.path.normpath(data_dir)
|
| 1340 |
+
config_kwargs_to_add_to_suffix["data_dir"] = data_dir
|
| 1341 |
+
if config_kwargs_to_add_to_suffix:
|
| 1342 |
+
# we don't care about the order of the kwargs
|
| 1343 |
+
config_kwargs_to_add_to_suffix = {
|
| 1344 |
+
k: config_kwargs_to_add_to_suffix[k] for k in sorted(config_kwargs_to_add_to_suffix)
|
| 1345 |
+
}
|
| 1346 |
+
if all(isinstance(v, (str, bool, int, float)) for v in config_kwargs_to_add_to_suffix.values()):
|
| 1347 |
+
suffix = ",".join(
|
| 1348 |
+
str(k) + "=" + urllib.parse.quote_plus(str(v)) for k, v in config_kwargs_to_add_to_suffix.items()
|
| 1349 |
+
)
|
| 1350 |
+
if len(suffix) > 32: # hash if too long
|
| 1351 |
+
suffix = Hasher.hash(config_kwargs_to_add_to_suffix)
|
| 1352 |
+
else:
|
| 1353 |
+
suffix = Hasher.hash(config_kwargs_to_add_to_suffix)
|
| 1354 |
+
|
| 1355 |
+
if custom_features is not None:
|
| 1356 |
+
m = Hasher()
|
| 1357 |
+
if suffix:
|
| 1358 |
+
m.update(suffix)
|
| 1359 |
+
m.update(custom_features)
|
| 1360 |
+
suffix = m.hexdigest()
|
| 1361 |
+
|
| 1362 |
+
if suffix:
|
| 1363 |
+
config_id = self.name + "-" + suffix
|
| 1364 |
+
if len(config_id) > config.MAX_DATASET_CONFIG_ID_READABLE_LENGTH:
|
| 1365 |
+
config_id = self.name + "-" + Hasher.hash(suffix)
|
| 1366 |
+
return config_id
|
| 1367 |
+
return self.name
|
| 1368 |
+
|
| 1369 |
+
|
| 1370 |
+
class ForNet(datasets.GeneratorBasedBuilder):
|
| 1371 |
+
"""ForNet dataset."""
|
| 1372 |
+
|
| 1373 |
+
def __init__(self, *args, **kwargs):
|
| 1374 |
+
"""Initialize the ForNet Builder."""
|
| 1375 |
+
super().__init__(*args, **kwargs)
|
| 1376 |
+
self.cls_to_idx_locs = {}
|
| 1377 |
+
|
| 1378 |
+
BUILDER_CONFIGS = [
|
| 1379 |
+
ForNetConfig(
|
| 1380 |
+
name="fornet",
|
| 1381 |
+
version=datasets.Version("1.0.0", ""),
|
| 1382 |
+
description="ForNet dataset",
|
| 1383 |
+
background_combination="all",
|
| 1384 |
+
fg_scale_jitter=0.3,
|
| 1385 |
+
pruning_ratio=0.8,
|
| 1386 |
+
fg_size_mode="range",
|
| 1387 |
+
fg_bates_n=1,
|
| 1388 |
+
mask_smoothing_sigma=4.0,
|
| 1389 |
+
rel_jut_out=0.0,
|
| 1390 |
+
orig_img_prob=0.0,
|
| 1391 |
+
)
|
| 1392 |
+
]
|
| 1393 |
+
|
| 1394 |
+
DEFAULT_WRITER_BATCH_SIZE = 1000
|
| 1395 |
+
|
| 1396 |
+
def _info(self):
|
| 1397 |
+
return datasets.DatasetInfo(
|
| 1398 |
+
description=_DESCRIPTION,
|
| 1399 |
+
features=datasets.Features(
|
| 1400 |
+
{
|
| 1401 |
+
"path": datasets.Value("string"),
|
| 1402 |
+
"bg": datasets.features.Image(),
|
| 1403 |
+
"fg": datasets.features.Image(),
|
| 1404 |
+
"in": datasets.features.Image(),
|
| 1405 |
+
"label": datasets.features.ClassLabel(names=list(IMAGENET2012_CLASSES.values())),
|
| 1406 |
+
"fg/bg_area": datasets.Value("float"),
|
| 1407 |
+
"bg_rat_idx_file": datasets.Value("string"),
|
| 1408 |
+
}
|
| 1409 |
+
),
|
| 1410 |
+
supervised_keys=("image", "label"),
|
| 1411 |
+
homepage=_HOMEPAGE,
|
| 1412 |
+
citation=_CITATION,
|
| 1413 |
+
)
|
| 1414 |
+
|
| 1415 |
+
def _split_generators(self, dl_manager: datasets.DownloadManager):
|
| 1416 |
+
urls_to_download = _CONST_URLS + _PATCH_URLS
|
| 1417 |
+
dl_paths = dl_manager.download(urls_to_download)
|
| 1418 |
+
|
| 1419 |
+
train_re = re.compile(r".*/train_(\d+)\.zip$")
|
| 1420 |
+
val_re = re.compile(r".*/val\.zip$")
|
| 1421 |
+
|
| 1422 |
+
train_patches = [f for f in dl_paths if train_re.match(f)]
|
| 1423 |
+
val_patches = [f for f in dl_paths if val_re.match(f)]
|
| 1424 |
+
|
| 1425 |
+
hf_train_indices = [f for f in dl_paths if f.endswith("hf_train_indices.json")][0]
|
| 1426 |
+
hf_val_indices = [f for f in dl_paths if f.endswith("hf_val_indices.json")][0]
|
| 1427 |
+
|
| 1428 |
+
cls_to_idx_locs = {
|
| 1429 |
+
"train": hf_train_indices.replace("hf_train_indices", "train_cls_to_idx"),
|
| 1430 |
+
"val": hf_val_indices.replace("hf_val_indices", "val_cls_to_idx"),
|
| 1431 |
+
}
|
| 1432 |
+
|
| 1433 |
+
fg_bg_ratios = [
|
| 1434 |
+
[f for f in dl_paths if f.endswith(f"fg_bg_ratios_{part}.json")][0] for part in ["train", "val"]
|
| 1435 |
+
]
|
| 1436 |
+
|
| 1437 |
+
return [
|
| 1438 |
+
datasets.SplitGenerator(
|
| 1439 |
+
name=datasets.Split.TRAIN,
|
| 1440 |
+
gen_kwargs={
|
| 1441 |
+
"patch_files": train_patches,
|
| 1442 |
+
"split": "train",
|
| 1443 |
+
"hf_indices": hf_train_indices,
|
| 1444 |
+
"cls_to_idx_loc": cls_to_idx_locs["train"],
|
| 1445 |
+
"fg_bg_ratios": fg_bg_ratios[0],
|
| 1446 |
+
},
|
| 1447 |
+
),
|
| 1448 |
+
datasets.SplitGenerator(
|
| 1449 |
+
name=datasets.Split.VALIDATION,
|
| 1450 |
+
gen_kwargs={
|
| 1451 |
+
"patch_files": val_patches,
|
| 1452 |
+
"split": "val",
|
| 1453 |
+
"hf_indices": hf_val_indices,
|
| 1454 |
+
"cls_to_idx_loc": cls_to_idx_locs["val"],
|
| 1455 |
+
"fg_bg_ratios": fg_bg_ratios[1],
|
| 1456 |
+
},
|
| 1457 |
+
),
|
| 1458 |
+
]
|
| 1459 |
+
|
| 1460 |
+
def _generate_examples(
|
| 1461 |
+
self, patch_files, split, hf_indices, cls_to_idx_loc, fg_bg_ratios
|
| 1462 |
+
): # TODO: Parallelize this with multiple tar extractor processes and also multiple recombiner processes. Iterate through imagenet in main thread only, I guess...
|
| 1463 |
+
logger.info(f"Generating examples from {len(patch_files)} patch files")
|
| 1464 |
+
logger.info("Opening files")
|
| 1465 |
+
class_to_zipfile = {}
|
| 1466 |
+
for f in patch_files:
|
| 1467 |
+
with zipfile.ZipFile(f, "r") as zf:
|
| 1468 |
+
for name in zf.namelist():
|
| 1469 |
+
if name.endswith(".pkl") or name.endswith(".pkl.gz"):
|
| 1470 |
+
class_to_zipfile[name.split("/")[-2]] = f
|
| 1471 |
+
file_ending = "pkl" if name.endswith(".pkl") else "pkl.gz"
|
| 1472 |
+
name_start = "/".join(name.split("/")[:-2])
|
| 1473 |
+
if len(name_start) > 0:
|
| 1474 |
+
name_start += "/"
|
| 1475 |
+
logger.info(f"Loading extra information: {hf_indices}, {fg_bg_ratios}")
|
| 1476 |
+
with open(hf_indices, "r") as f:
|
| 1477 |
+
path_to_in_idx = json.load(f)
|
| 1478 |
+
idx_to_path = {v: k for k, v in path_to_in_idx.items()}
|
| 1479 |
+
# print("idx_to_path", list(idx_to_path.items())[:5])
|
| 1480 |
+
with open(fg_bg_ratios, "r") as f:
|
| 1481 |
+
fg_bg_ratios = json.load(f)
|
| 1482 |
+
fg_bg_ratios = {"/".join(k.split("/")[-2:]).split(".")[0]: v for k, v in fg_bg_ratios.items()}
|
| 1483 |
+
# print("fg_bg_ratios", list(fg_bg_ratios.items())[:5])
|
| 1484 |
+
|
| 1485 |
+
logger.info("Starting extraction with ImageNet")
|
| 1486 |
+
foraug_idx = 0
|
| 1487 |
+
|
| 1488 |
+
manager = multiprocessing.Manager()
|
| 1489 |
+
num_workers = multiprocessing.cpu_count()
|
| 1490 |
+
if os.environ.get("MAX_WORKERS", None):
|
| 1491 |
+
num_workers = int(os.environ["MAX_WORKERS"])
|
| 1492 |
+
in_queue = manager.Queue(maxsize=4 * num_workers)
|
| 1493 |
+
ret_queue = manager.Queue(maxsize=4 * num_workers)
|
| 1494 |
+
comm_dict = manager.dict()
|
| 1495 |
+
comm_dict["running"] = True
|
| 1496 |
+
comm_dict["n_errors"] = 0
|
| 1497 |
+
|
| 1498 |
+
in_proc = multiprocessing.Process(target=_in_iterator, args=(in_queue, split))
|
| 1499 |
+
in_proc.start()
|
| 1500 |
+
zip_procs = [
|
| 1501 |
+
multiprocessing.Process(
|
| 1502 |
+
target=_zip_loader,
|
| 1503 |
+
args=(
|
| 1504 |
+
in_queue,
|
| 1505 |
+
ret_queue,
|
| 1506 |
+
comm_dict,
|
| 1507 |
+
patch_files,
|
| 1508 |
+
class_to_zipfile,
|
| 1509 |
+
name_start,
|
| 1510 |
+
file_ending,
|
| 1511 |
+
idx_to_path,
|
| 1512 |
+
fg_bg_ratios,
|
| 1513 |
+
),
|
| 1514 |
+
)
|
| 1515 |
+
for _ in range(num_workers)
|
| 1516 |
+
]
|
| 1517 |
+
for proc in zip_procs:
|
| 1518 |
+
proc.start()
|
| 1519 |
+
|
| 1520 |
+
last_errors = 0
|
| 1521 |
+
cls_to_idx = {}
|
| 1522 |
+
while comm_dict["running"]:
|
| 1523 |
+
if not ret_queue.empty():
|
| 1524 |
+
data = ret_queue.get()
|
| 1525 |
+
in_cls = data["path"].split("/")[0]
|
| 1526 |
+
if in_cls not in cls_to_idx:
|
| 1527 |
+
cls_to_idx[in_cls] = []
|
| 1528 |
+
cls_to_idx[in_cls].append((foraug_idx, data["fg/bg_area"]))
|
| 1529 |
+
if foraug_idx == 0:
|
| 1530 |
+
data["bg_rat_idx_file"] = cls_to_idx_loc
|
| 1531 |
+
yield foraug_idx, data
|
| 1532 |
+
foraug_idx += 1
|
| 1533 |
+
|
| 1534 |
+
if comm_dict["n_errors"] > last_errors:
|
| 1535 |
+
last_errors = comm_dict["n_errors"]
|
| 1536 |
+
tqdm.write(
|
| 1537 |
+
f"@step {foraug_idx}: errors {last_errors}; error rate {last_errors / foraug_idx:.2%} (expected {6_610 / 1_274_227:.2%})"
|
| 1538 |
+
)
|
| 1539 |
+
|
| 1540 |
+
if ret_queue.empty() and in_queue.empty() and not in_proc.is_alive():
|
| 1541 |
+
comm_dict["running"] = False
|
| 1542 |
+
tqdm.write("Finished imagenet iteration; waiting for zip loaders to finish")
|
| 1543 |
+
|
| 1544 |
+
in_proc.join()
|
| 1545 |
+
for proc in zip_procs:
|
| 1546 |
+
proc.join()
|
| 1547 |
+
|
| 1548 |
+
# tqdm.write("Finished all processes")
|
| 1549 |
+
while not ret_queue.empty():
|
| 1550 |
+
data = ret_queue.get()
|
| 1551 |
+
in_cls = data["path"].split("/")[0]
|
| 1552 |
+
if in_cls not in cls_to_idx:
|
| 1553 |
+
cls_to_idx[in_cls] = []
|
| 1554 |
+
cls_to_idx[in_cls].append(foraug_idx)
|
| 1555 |
+
yield foraug_idx, data
|
| 1556 |
+
foraug_idx += 1
|
| 1557 |
+
# tqdm.write("Done")
|
| 1558 |
+
with open(cls_to_idx_loc, "w") as f:
|
| 1559 |
+
json.dump(cls_to_idx, f)
|
| 1560 |
+
|
| 1561 |
+
def _as_streaming_dataset_single(self, *args, **kwargs):
|
| 1562 |
+
raise NotImplementedError("ForNet does not support streaming datasets")
|
| 1563 |
+
|
| 1564 |
+
def _as_dataset(self, split=datasets.Split.TRAIN, in_memory=False):
|
| 1565 |
+
"""Constructs a `Dataset`.
|
| 1566 |
+
|
| 1567 |
+
This is the internal implementation to overwrite called when user calls
|
| 1568 |
+
`as_dataset`. It should read the pre-processed datasets files and generate
|
| 1569 |
+
the `Dataset` object.
|
| 1570 |
+
|
| 1571 |
+
Args:
|
| 1572 |
+
split (`datasets.Split`):
|
| 1573 |
+
which subset of the data to read.
|
| 1574 |
+
in_memory (`bool`, defaults to `False`):
|
| 1575 |
+
Whether to copy the data in-memory.
|
| 1576 |
+
|
| 1577 |
+
Returns:
|
| 1578 |
+
`Dataset`
|
| 1579 |
+
"""
|
| 1580 |
+
cache_dir = self._fs._strip_protocol(self._output_dir)
|
| 1581 |
+
dataset_name = self.dataset_name
|
| 1582 |
+
if self._check_legacy_cache():
|
| 1583 |
+
dataset_name = self.name
|
| 1584 |
+
dataset_kwargs = ArrowReader(cache_dir, self.info).read(
|
| 1585 |
+
name=dataset_name,
|
| 1586 |
+
instructions=split,
|
| 1587 |
+
split_infos=self.info.splits.values(),
|
| 1588 |
+
in_memory=in_memory,
|
| 1589 |
+
)
|
| 1590 |
+
|
| 1591 |
+
fingerprint = self._get_dataset_fingerprint(split)
|
| 1592 |
+
# print("config", self.config)
|
| 1593 |
+
# print("rel data dir", self._relative_data_dir())
|
| 1594 |
+
# print("split", str(split))
|
| 1595 |
+
# print("fingerprint", fingerprint)
|
| 1596 |
+
|
| 1597 |
+
splitname = str(split)
|
| 1598 |
+
if splitname == "validation":
|
| 1599 |
+
splitname = "val"
|
| 1600 |
+
|
| 1601 |
+
return RecombineDataset(
|
| 1602 |
+
fingerprint=fingerprint,
|
| 1603 |
+
background_combination=self.config.background_combination,
|
| 1604 |
+
fg_scale_jitter=self.config.fg_scale_jitter,
|
| 1605 |
+
pruning_ratio=self.config.pruning_ratio,
|
| 1606 |
+
fg_size_mode=self.config.fg_size_mode,
|
| 1607 |
+
fg_bates_n=self.config.fg_bates_n,
|
| 1608 |
+
mask_smoothing_sigma=self.config.mask_smoothing_sigma,
|
| 1609 |
+
rel_jut_out=self.config.rel_jut_out,
|
| 1610 |
+
orig_img_prob=self.config.orig_img_prob,
|
| 1611 |
+
**dataset_kwargs,
|
| 1612 |
+
)
|
| 1613 |
+
|
| 1614 |
+
def _create_builder_config(self, config_name=None, custom_features=None, **config_kwargs):
|
| 1615 |
+
config_hash_kwargs = {k: v for k, v in config_kwargs.items() if k not in _CONFIG_HASH_IGNORE_KWARGS}
|
| 1616 |
+
builder_config, config_id = super()._create_builder_config(config_name, custom_features, **config_hash_kwargs)
|
| 1617 |
+
for k in _CONFIG_HASH_IGNORE_KWARGS:
|
| 1618 |
+
if k in config_kwargs:
|
| 1619 |
+
setattr(builder_config, k, config_kwargs[k])
|
| 1620 |
+
return builder_config, config_id
|
| 1621 |
+
|
| 1622 |
+
|
| 1623 |
+
def _in_iterator(in_queue, split):
|
| 1624 |
+
if split == "val":
|
| 1625 |
+
split = "validation"
|
| 1626 |
+
imagenet = datasets.load_dataset("ILSVRC/imagenet-1k", split=split)
|
| 1627 |
+
for idx, ex in enumerate(imagenet):
|
| 1628 |
+
in_queue.put((idx, ex["image"]))
|
| 1629 |
+
|
| 1630 |
+
|
| 1631 |
+
def _zip_loader(
|
| 1632 |
+
in_queue, ret_queue, comm_dict, patch_files, class_to_zipfile, name_start, file_ending, idx_to_path, fg_bg_ratios
|
| 1633 |
+
):
|
| 1634 |
+
while comm_dict["running"]:
|
| 1635 |
+
if not in_queue.empty():
|
| 1636 |
+
try:
|
| 1637 |
+
in_idx, in_img = in_queue.get(block=False)
|
| 1638 |
+
except queue.Empty:
|
| 1639 |
+
continue
|
| 1640 |
+
patch_name = idx_to_path[in_idx]
|
| 1641 |
+
in_class, in_file_name = patch_name.split("/")
|
| 1642 |
+
|
| 1643 |
+
try:
|
| 1644 |
+
with zipfile.ZipFile(class_to_zipfile[in_class], "r") as zf, (
|
| 1645 |
+
zf.open(f"{name_start}{patch_name}.{file_ending}", "r")
|
| 1646 |
+
if file_ending == "pkl"
|
| 1647 |
+
else gzip.GzipFile(fileobj=zf.open(f"{name_start}{patch_name}.{file_ending}", "r"), mode="r")
|
| 1648 |
+
) as pklf:
|
| 1649 |
+
patch_data = pickle.load(pklf)
|
| 1650 |
+
except KeyError:
|
| 1651 |
+
comm_dict["n_errors"] += 1
|
| 1652 |
+
continue
|
| 1653 |
+
|
| 1654 |
+
in_img = in_img.convert("RGB")
|
| 1655 |
+
|
| 1656 |
+
if "bg_diff" in patch_data:
|
| 1657 |
+
if in_img.size != (patch_data["bg_diff"].shape[1], patch_data["bg_diff"].shape[0]):
|
| 1658 |
+
in_img = in_img.resize((patch_data["bg_diff"].shape[1], patch_data["bg_diff"].shape[0]))
|
| 1659 |
+
else:
|
| 1660 |
+
max_size = max(in_img.size)
|
| 1661 |
+
if max_size > 512:
|
| 1662 |
+
goal_size = (round(in_img.size[0] * 512 / max_size), round(in_img.size[1] * 512 / max_size))
|
| 1663 |
+
in_img = in_img.resize(goal_size)
|
| 1664 |
+
|
| 1665 |
+
in_img = np.array(in_img)
|
| 1666 |
+
|
| 1667 |
+
if "bg_diff" in patch_data:
|
| 1668 |
+
bg_diff = patch_data["bg_diff"]
|
| 1669 |
+
bg_img = in_img.astype(np.int64) + bg_diff
|
| 1670 |
+
else:
|
| 1671 |
+
bg_img = None
|
| 1672 |
+
|
| 1673 |
+
if "fg_mask" in patch_data:
|
| 1674 |
+
x_offs, y_offs = patch_data["fg_off"]
|
| 1675 |
+
fg_mask = patch_data["fg_mask"]
|
| 1676 |
+
|
| 1677 |
+
fg_crop = in_img[y_offs : y_offs + fg_mask.shape[0], x_offs : x_offs + fg_mask.shape[1]]
|
| 1678 |
+
fg_img = np.concatenate([fg_crop, fg_mask * 255], axis=-1).clip(0, 255).astype(np.uint8)
|
| 1679 |
+
else:
|
| 1680 |
+
fg_img = None
|
| 1681 |
+
ret_queue.put(
|
| 1682 |
+
{
|
| 1683 |
+
"path": patch_name,
|
| 1684 |
+
"bg": bg_img,
|
| 1685 |
+
"fg": fg_img,
|
| 1686 |
+
"label": IMAGENET2012_CLASSES[in_class],
|
| 1687 |
+
"in": in_img,
|
| 1688 |
+
"fg/bg_area": fg_bg_ratios[patch_name],
|
| 1689 |
+
}
|
| 1690 |
+
)
|