songincv / gmphd-ogm_tracker Goto Github PK

Hi Young-min,
It seems that you have very good experience in MOT.
I have run my detector on embedded system so I just have detections every 120ms(frame rate is 30fps)...(by GPU)
now I need MOT to run on CPU simultaneously.
And I need track to run between frames.
What do you suggest?

thanks

VISUALIZATION_MAIN_ON 을 1로 변경후 디버깅없이 실행(Ctrl + F5)할 때에 오픈시피 오류가 납니다.

0으로 했을 시에는 오류가 나지않습니다.

오류 화면 캡쳐합니다. 혹시 짐작가는 해결방법이 있을까요? (demo_....cpp 에서 seqList만 제 이미지 경로로변경을하고 paramsFile은 변경하지 않긴 했습니다. 이 것이 오류의 원인일 수도 있 을까요? 또한 'qaramsFile`이 가리키는 파일들은 무엇을 의미하나요?)

+추가

에러부분을 찾으려고 디버깅모드(F5)로 해보니 에러가 뜨긴하지만 실행이 되긴하네요... 무슨상황일까요

주어진 코드와 파라미터 값들을 사용하여 MOT17-FRCNN을 테스트 하였을 때, GMPHD_OGM::MinimizeGroupCost 함수의 this->tracksbyID[groupRects[v].id].pop_back(); 부분에서 vector out of range error가 확인되었습니다. pop_back()을 수행하는 과정에서 size가 0이 되면서 오류가 발생한 것 같으니 확인 부탁드립니다.

When i tested "MOT17-FRCNN" sequences with the given code and parameters, the 'vector out of range error' occurs in this->tracksbyID[groupRects[v].id].pop_back(); in 'GMPHD_OGM::MinimizeGroupCost' function. It seems to be an error in the process of performing pop_back(), as the size became zero.

In the MinimizeGroupCost method, there is a loop on v, and inside that a loop on fr. Looks like fr indicates in part how many items have been processed and removed from the gid_track list. It's very confusing and uncommented code. It looks like elements are being removed from "gid_track" and added to "track". The problem is that there is a "gid_track[0].pop_back()" to remove the last item. But right after that, a check is made against the last element at (tSize - 1 - fr). But, this element position was just removed. What exactly is the intent of this loop on "fr" inside a loop on "v" ?

It looks like you fixed this in a previous issue, but brought it back more recently in an optimization check-in. Was this a mistake?