Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Detect new object and Keep tracking of old obejct #3

Open
Riser6 opened this issue Aug 21, 2024 · 3 comments
Open

Detect new object and Keep tracking of old obejct #3

Riser6 opened this issue Aug 21, 2024 · 3 comments

Comments

@Riser6
Copy link

Riser6 commented Aug 21, 2024

Hi,
Thanks for this very wonderful and useful project. I wondered if it would be possible to accept streaming video as input, constantly detect new objects and track them, while keeping track of old ones (even if they disappear for a while due to occlusion).
Thanks in advance!

@Gy920
Copy link
Owner

Gy920 commented Aug 23, 2024

As mentioned in #2, the current pipeline requires resetting the tracker to add new objects. This method can lead to tracking loss if the old object is occluded and its prompt cannot be obtained in the current frame.

A better approach would be to preserve the memory of past feature frames output_dict["non_cond_frame_outputs"], containing the old object when resetting the tracker, specifically by modifying this part . This way, objects that are occluded for a long time can still be tracked.

@Riser6
Copy link
Author

Riser6 commented Aug 23, 2024

As mentioned in #2, the current pipeline requires resetting the tracker to add new objects. This method can lead to tracking loss if the old object is occluded and its prompt cannot be obtained in the current frame.

A better approach would be to preserve the memory of past feature frames output_dict["non_cond_frame_outputs"], containing the old object when resetting the tracker, specifically by modifying this part . This way, objects that are occluded for a long time can still be tracked.

Thank you very much for your detailed reply. I will check this issue according to your opinion.

@Greywan
Copy link

Greywan commented Sep 23, 2024

As mentioned in #2, the current pipeline requires resetting the tracker to add new objects. This method can lead to tracking loss if the old object is occluded and its prompt cannot be obtained in the current frame.
A better approach would be to preserve the memory of past feature frames output_dict["non_cond_frame_outputs"], containing the old object when resetting the tracker, specifically by modifying this part . This way, objects that are occluded for a long time can still be tracked.

Thank you very much for your detailed reply. I will check this issue according to your opinion.

May I ask if you've achieved this effect?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

4 participants
@Riser6 @Greywan @Gy920 and others