Skip to content

dnn/video: TrackerVit crashes when target is lost #25760

@fengyuentau

Description

@fengyuentau

System Information

Latest OpenCV

Detailed description

Origin opencv/opencv_zoo#261

When the target is lost, TrackerVit records the wrong bbox and therefore the next bbox will always be wrong and become larger and larger until out-of-memory error.

Steps to reproduce

  1. Install latest opencv-python and get latest opencv_zoo.
  2. cd $opencv_zoo/models/object_tracking_vittrack
  3. python demo.py
  4. Select an object for tracking, press enter to enter tracking process.
  5. Use something thick to cover the camera to make frames all black. Inference speed goes slower and slower and demo will crash due to OOM.

Issue submission checklist

  • I report the issue, it's not a question
  • I checked the problem with documentation, FAQ, open issues, forum.opencv.org, Stack Overflow, etc and have not found any solution
  • I updated to the latest OpenCV version and the issue is still there
  • There is reproducer code and related data files (videos, images, onnx, etc)

Metadata

Metadata

Assignees

Type

No type

Projects

No projects

Milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions