search

alanlukezic / d3s

D3S - Discriminative Single Shot Segmentation Tracker (CVPR 2020)
265 stars 56 forks source link

How to training for video object segmentation #19

Open limoran233 opened 4 years ago

limoran233 commented 4 years ago

hello，thanks for your excellent job! I want to know whether the code can be trained on YouTube-VOS and then evaluate on DAVIS for VOS task，and how to implement it？Looking forward to your reply.

alanlukezic commented 4 years ago

The pre-trained model which was used for reporting results in the paper and is available online has already been trained on YT-VOS. If you want to use it on DAVIS you have to create your own script which initializes the tracker on the first frame and then tracks it on every other frame. Current implementation can save the predicted segmentation masks if params.masks_save_path in the pytracking/parameter/segm/default_params.py is set to the directory where you want the masks to be saved.