search

Dawn-LX / OpenVoc-VidVRD

Official code for the ICLR2023 paper Compositional Prompt Tuning with Motion Cues for Open-vocabulary Video Relation Detection
37 stars 1 forks source link

How to extract RoI features #3

Closed QiueY514 closed 11 months ago

QiueY514 commented 11 months ago

Hi, thank you for your excellent work! The released RoI features have been averaged of all the bounding boxes along the temporal axis of the tracklets. So could you release the RoI features for each bounding box without averaging? Thank you very much.

Dawn-LX commented 11 months ago

we did not store the frame-level RoI features. But we can release the code for extracting trackets' RoI features. Maybe 3~5 days later.

QiueY514 commented 11 months ago

we did not store the frame-level RoI features. But we can release the code for extracting trackets' RoI features. Maybe 3~5 days later.

Thank you for your generous sharing!

Dawn-LX commented 11 months ago

we did not store the frame-level RoI features. But we can release the code for extracting trackets' RoI features. Maybe 3~5 days later.

Thank you for your generous sharing!

Hi, you can refer to this repo VidSGG-TrajDataPrepare for traj data preparing. It include 3 stages: 1)detect bbox --> 2)object tracking --> 3)extract RoI features.

QiueY514 commented 11 months ago

we did not store the frame-level RoI features. But we can release the code for extracting trackets' RoI features. Maybe 3~5 days later.

Thank you for your generous sharing!

Hi, you can refer to this repo VidSGG-TrajDataPrepare for traj data preparing. It include 3 stages: 1)detect bbox --> 2)object tracking --> 3)extract RoI features.

Thank you for your shared code! It helps me a lot.