foolwood / SiamMask

[CVPR2019] Fast Online Object Tracking and Segmentation: A Unifying Approach
http://www.robots.ox.ac.uk/~qwang/SiamMask
MIT License
3.47k stars 816 forks source link

Train this network #52

Open qintaohu opened 5 years ago

qintaohu commented 5 years ago

Sorry,if I want to train this network, can i use only 1 dataset from Youtube-VOS, COCO, ImageNet-DET and ImageNet-VID, or all. thanks!

TeslaHua commented 5 years ago

@ustcerman You can do this, you need to comment out the other datasets in config.json.

xieyunjiao commented 4 years ago

Excuse me,have you successfully changed the data set and trained? If you succeed, can you tell me how to do that?

Damon2019 commented 4 years ago

Excuse me,have you successfully changed the data set and trained? If you succeed, can you tell me how to do that?

hi,i also meet this problem.Can you tell me how to do?thanks

asuleyman commented 3 years ago

Have you find an answer? I do not think it is possible to use only one dataset because each dataset has a specific task (detection, segmentation)