i change detector ,use FCOS instead of YOlO, The effect of pedestrian detection is good, i want to Detect a frame to track 10 frames,it seem some trouble,can you show your code for real time processing

[huangzicheng]

[ZQPei]

what do you mean by Detecting a frame to track 10 frames? If you want to track 10 frames, then detect 10 frames, because you need feed the detection results of each frame to the tracker to update the bounding box info. Otherwise, you can set the frame interval to 10 for both detector and tracker to speed up. Make sure the detector and tracker shall process simultaneously.

[huangzicheng]

I want to speed up the real-time processing speed, detect a frame of about 5-10fps, track a frame of about 80fps, there are two ways of tracking, track_with_detect and track_without_detect, only an interval of 10 frames for detection as a track correction

[ZQPei]

Set an interval to 10 to see if it works. The bottleneck is the detection speed. Rather than set an interval, it is better to adopt a lightweight detection model or accerate your model by pruning or quantization.

[Zzzcg]

I want to speed up the real-time processing speed, detect a frame of about 5-10fps, track a frame of about 80fps, there are two ways of tracking, track_with_detect and track_without_detect, only an interval of 10 frames for detection as a track correction

hi，would you like to tell me how to set interval? thanks a lot

[xiaozhi2015]

You can set interval with para "frame_interval" in yolo3_deepsort.py. @Zzzcg About "Detecting a frame to track 10 frames", if you detect every frame, then why do we need to track, the detect result of every frame is enough. Detect result as a goal, then track the goal with only tracking algorithm, that's what Track is all about @ZQPei Have you realized "Detecting a frame to track 10 frames" bro? @huangzicheng