Could you release the code related to detection DA?

valeoai / ADVENT

Adversarial Entropy Minimization for Domain Adaptation in Semantic Segmentation

https://arxiv.org/abs/1811.12833

Other

382 stars 70 forks source link

Could you release the code related to detection DA? #14

Closed dlsrbgg33 closed 4 years ago

dlsrbgg33 commented 4 years ago

Hi, Thank you for your great work and contribution. As a title, could you release the SSD code with your DA approach? Thank you.

tuanhungvu commented 4 years ago

Hello @dlsrbgg33 , sorry for the late replay. Thanks for your interest in our work. Currently we don't have a concrete plan to release the detection part, mainly due to our lack of time. FYI, the detection part is built upon this repo https://github.com/amdegroot/ssd.pytorch . Best, T-H

andgitchang commented 4 years ago

@tuanhungvu I would like to know the advent detection used (1) only the classification header of SSD to calculate entropy maps or (2) both classification and regression headers to adapt? It seems that the classification header for entropy maps is more reasonable, said 80 classes of COCO, where 4 classes bounding box (xyxy) of the regression header calculating entropy maps is a bit of weird. Thanks in advance.