updated the description
Browse files
README.md
CHANGED
@@ -5,17 +5,17 @@ tags:
|
|
5 |
- Architecture
|
6 |
---
|
7 |
|
8 |
-
|
9 |
|
10 |
-
|
11 |
|
12 |
-
##
|
13 |
|
14 |
-
|
15 |
|
16 |
-
##
|
17 |
|
18 |
-
|
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 |
|