CNNs from Scratch

Build convolutional networks one block at a time — from raw conv to skip connections and squeeze-excitation.

1. 1. ○ Implement 1D Convolution
2. 2. ○ Implement Conv2d
3. 3. ○ Implement Max Pooling 1D
4. 4. ○ Implement Average Pooling 1D
5. 5. ○ Depthwise Separable Convolution
6. 6. ○ Transposed Conv2d
7. 7. ○ Skip Connection Block
8. 8. ○ Squeeze-and-Excitation Block
9. 9. ○ Simple CNN
10. 10. ○ Train Image Classifier End-to-End
11. 11. ○ Patch Embedding with CLS Token
12. 12. ○ ViT Encoder Block
13. 13. ○ ViT Classification Forward
14. 14. ○ Train ViT Classifier End-to-End
15. 15. ○ ViT with Mixup Augmentation