escaping bbox

[ksh1983]

When the face moves or under certain conditions, the bounding box randomly shrinks and grows, eventually disappearing off the screen. The face is not detected again until it reappears on the screen. This issue is more likely to occur when the face moves quickly or when there are two or more people in the webcam frame. Additionally, the smaller the bounding box and the further the detection distance, the more frequently this issue occurs. Can you help me understand why this is happening?

[vectornguyen76]

The reason the above problem occurred was due to my ineffective use of multi threads. The idea to solve this problem is to combine recognition and detection into 2 separate applications, and use redis to communicate. @ksh1983

[ksh1983]

2 separate applications, and use redis to communicate <<<I don't understand, can you show me an example?

[vectornguyen76]

you can run 2 containers for detection and recognition. And then you use redis to push frame detected from detection to recognition container

[ksh1983]

I don't think it's a multi-thread problem, so please check it. It's a video I shot myself, and if I use cuda, it comes out more than 60fps, but in the video, the fps comes out low due to the capture setting. If you move your face quickly like that or if more than two people appear on the screen at the same time, that happens. 2024년 6월 26일 (수) 오전 1:22, Vector Nguyễn ***@***.***>님이 작성:

…

you can run 2 containers for detection and recognition. And then you use redis to push frame detected from detection to recognition container — Reply to this email directly, view it on GitHub <#26 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AKCPASMM66AN4ZSKQMVAWETZJGKLDAVCNFSM6AAAAABJSAODZOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDCOBZGM4TKOJWGE> . You are receiving this because you were mentioned.Message ID: ***@***.***>