Timm: - Training (Fine-tuning) all branches with full loss. Use pre-trained SWIN (frozen), and pre-trained ```gt_decoder``` and ```sl_decoder``` (NOT frozen) Boris and Mari: - Generate occlusion targets and start pre-training for ```occ_decoder``` Janice: - Continue writing intermediate reports - E.g. Describe (3D) SWIN architecture
Timm:
gt_decoderandsl_decoder(NOT frozen)Boris and Mari:
occ_decoderJanice: