Weekly Summary

1. Reviewed FCN and ResNet architecture
2. Familiarized myself with mmsegmentation toolbox
3. Completed torchvision-style ResNet to mmseg-style ResNet model conversion
4. Done some lightweight testing using MoCo pre-trained ResNet-50 weight with FCN head

   To-do’s

• Reproduce fully-supervised ImageNet pre-training with FCN head finetuning on VOC segmentation benchmark
• Convert SimCLR (tensorflow model) to pytorch
• Test with SimCLR