harsha163 commited on
Commit
bb701de
·
1 Parent(s): aca7595

updated the description

Browse files
Files changed (1) hide show
  1. README.md +6 -6
README.md CHANGED
@@ -5,17 +5,17 @@ tags:
5
  - Architecture
6
  ---
7
 
8
- ## Model description
9
 
10
- More information needed
11
 
12
- ## Intended uses & limitations
13
 
14
- More information needed
15
 
16
- ## Training and evaluation data
17
 
18
- More information needed
19
 
20
  ## Training procedure
21
 
 
5
  - Architecture
6
  ---
7
 
8
+ # Tensorflow Keras implementation of : [Image classification with ConvMixer](https://keras.io/examples/vision/convmixer/)
9
 
10
+ The full credit goes to: [Sayak Paul](https://twitter.com/RisingSayak)
11
 
12
+ ## Short description:
13
 
14
+ ConvMixer is a simple model based on the ideas of representing an image as patches( used in ViT) and separating the mixing of Spatial and channel dimensions (used in MLP-Mixer). Unlike ViT and MLP-Mixer, they use only standard Convolution operations. The full paper is a submission to ICLR 22 and can be found [here](https://openreview.net/pdf?id=TVHS5Y4dNvM)
15
 
16
+ ## Model and Dataset used
17
 
18
+ The Dataset used here is CIFAR-10. The model is called ConvMixer-256/8 where 256 is the hidden dimension (the dimension of patches) and 8 is the depth(number of repetitions of ConvMix layers)
19
 
20
  ## Training procedure
21