Modified-RegNetX-200MF

This project uses PyTorch to modify, train and implement RegNetX-200MF for image classification of a modified variant of ImageNet. The dataset was created by down-sampling the original ImageNet images such that their short side is 64 pixels (while the other side is >= 64 pixels) and only 100 of the original 1000 classes were kept.

NOTE: This project was focused on understanding the different building block structures of a CNN for image classification, not to improve classification accuracy.

Modified Architecture

• Set stride = 1 (instead of stride = 2) in the stem
• Replace the first stride = 2 down-sampling building block in the original network by a stride = 1 normal building block
• The fully connected layer in the decoder outputs 100 classes instead of 1000 classes
• All of the other blocks in RegNetX-200MF stay the same

The original RegNetX-200MF takes in 3x224x224 input images and generates Nx7x7 feature maps before the decoder. This modified RegNetX-200MF will take in 3x56x56 input images (cropped from the provided data set) and generate Nx7x7 feature maps before the decoder.

Refer to cnn.pdf for more details

Training

Number of epochs = 125
Batch size = 512
Learning rate = linear warmup for 5 epochs followed by cosine decay for 120 epochs
Error criterion = softmax cross entropy
Optimizer = Adam
Weight decay (l2 norm) = 0

Final accuracy = 67.32%

Instructions on running the code

1. Go to Google Colaboratory: https://colab.research.google.com/notebooks/welcome.ipynb
2. File - New Python 3 notebook
3. Cut and paste this file into the cell (feel free to divide into multiple cells)
4. Runtime - Run all