Skip to content

Latest commit

 

History

History
19 lines (15 loc) · 894 Bytes

README.md

File metadata and controls

19 lines (15 loc) · 894 Bytes

Paper

Summary

  • Traditional tracking models are vulnerable to occlusion, which causes identity switches when an object (person) is occluded by another object.
  • This study uses two cameras to prevent identity switches for occluded objects.
  • This study is conducted in an indoor space and involves two people.
  • The idea is to enable tracking using two cameras without using a sophisticated object re-identification model such as deepsort.

Limitation

  • If an object is occluded and given a new ID, the issue is not resolved
  • Therefore I'm going to use two cameras to address the case that the correct ID changes another ID beacuse of occlusion.

References

Result Video