Amodal-Segmentation-Based-on-Visible-Region-Segmentation-and-Shape-Prior
Amodal-Segmentation-Based-on-Visible-Region-Segmentation-and-Shape-Prior copied to clipboard
--eval-only --resume doesn't load codebook?
Hi. I wanted to perform more detailed evaluations of an already trained model by running
train_net.py --eval-only --resume --config-file <my_config.yaml>
I found however that the system does not even attempt to load the learned codebook, and instead starts inference of the training dataset following with the clustering step to build it from scratch. This is of course impractical (it takes ~9h to cluster my dataset). Is there a "proper" way of performing evaluation only?
Alternatively, could you give me a hint, where in the code is this part of the model initialized, so I could monkeypatch a forced codebook load?