FOR Evaluation

1. Comment out line 9 and change the loss function from the minimal implementation in solov2_head.py
2. change the loss in line 310 after commenting out the init loss in solov2_head.py
3. Change np.bool to just bool in eval.py in line 188
4. Change the save path in eval.py and the save=True

FOR Training

1. in mask_feat_head.py, Make the following change: # feature_add_all_level += self.convs_all_levelsi feature_add_all_level = self.convs_all_levelsi + feature_add_all_level

Original REPO: https://github.com/OpenFirework/pytorch_solov2