Why segm_predictor() function need both train_feat_segm and test_feat_segm

alanlukezic / d3s

D3S - Discriminative Single Shot Segmentation Tracker (CVPR 2020)

265 stars 57 forks source link

Why segm_predictor() function need both train_feat_segm and test_feat_segm #17

Open wenching33 opened 3 years ago

wenching33 commented 3 years ago

Hi, I cant find segm_predictor() function implementation in this repository. Would please tell me where it is implemented ? and why it needs train_feat_segm as input argument ? Thanks

xiankgx commented 3 years ago

I'm thinking train_feat_segm are deep features extracted from the cropped image that we want to track which when combined with the train_masks will give hint to the model what are background and foreground pixels, while test_feat_segm are deep features extracted from the full image. This or the other way round.

xiankgx commented 3 years ago

Implementation is here.