Low evaluation of custom data sets and storage path problems

[wx-ikll007700]

Search before asking

• I have searched the Yolo Tracking issues and found no similar bug report.

Question

Hello, I am very interested in your project. I found my own data set and tried it. My data set has only one category. I trained my own detection weight (marked with gt file). After I ran val.py, I found that the results of bytetrack were all 0, and the results of deepocsort and botsort were the same, and they were very low. I felt that the detector did not work, but both my folders were generated.

[wx-ikll007700]

According to the results, my gt file should have been detected, but my detector doesn't seem to have detected my data.When I use bytetrack, the results are all 0, and the results of deepocsort and botsort are the same but extremely low.

[lmaple24327]

Hello, at present, I am also using YOLOv8 and BOTSORT for target tracking, and I also use my own custom data set, but the evaluation effect is also very poor. May I ask how do you use OCSORT? Do you use this program with many warehouses?

根据结果，我的 gt 文件应该已经被检测到，但我的检测器似乎没有检测到我的数据。当我使用 bytetrack 时，结果都是 0，deepocsort 和 botsort 的结果相同但极低。

[wx-ikll007700]

Hello, at present, I am also using YOLOv8 and BOTSORT for target tracking, and I also use my own custom data set, but the evaluation effect is also very poor. May I ask how do you use OCSORT? Do you use this program with many warehouses?

根据结果，我的 gt 文件应该已经被检测到，但我的检测器似乎没有检测到我的数据。当我使用 bytetrack 时，结果都是 0，deepocsort 和 botsort 的结果相同但极低。

Isn't this boxmot integrated with many tracking algorithms, and it only needs to be selected in the instruction, but my effect is very poor. It seems that I didn't put the path of the file right. Did you deploy it with this boxmot?

[github-actions]

👋 Hello, this issue has been automatically marked as stale because it has not had recent activity. Please note it will be closed if no further activity occurs.
Feel free to inform us of any other issues you discover or feature requests that come to mind in the future. Pull Requests (PRs) are also always welcomed!

[14790897]

I feel the same way. I think this document is very vague, and I don't know exactly how to run the evaluation. Also, I am evaluating on my own training dataset.

All sequences for  finished in 0.72 seconds

HOTA: -pedestrian                  HOTA      DetA      AssA      DetRe     DetPr     AssRe     AssPr     LocA      OWTA      HOTA(0)   LocA(0)   HOTALocA(0)
mot_particle                       36.257    30.554    44.278    36.564    44.857    48.719    60.939    67.89     40.074    81.835    45.642    37.351
COMBINED                           36.257    30.554    44.278    36.564    44.857    48.719    60.939    67.89     40.074    81.835    45.642    37.351

CLEAR: -pedestrian                 MOTA      MOTP      MODA      CLR_Re    CLR_Pr    MTR       PTR       MLR       sMOTA     CLR_TP    CLR_FN    CLR_FP    IDSW      MT        PT        ML        Frag
mot_particle                       -18.079   67.756    -18.016   31.748    38.949    3.6364    54.545    41.818    -28.316   1008      2167      1580      2         2         30        23        153
COMBINED                           -18.079   67.756    -18.016   31.748    38.949    3.6364    54.545    41.818    -28.316   1008      2167      1580      2         2         30        23        153

Identity: -pedestrian              IDF1      IDR       IDP       IDTP      IDFN      IDFP
mot_particle                       34.669    31.465    38.601    999       2176      1589
COMBINED                           34.669    31.465    38.601    999       2176      1589

Count: -pedestrian                 Dets      GT_Dets   IDs       GT_IDs
mot_particle                       2588      3175      55        55
COMBINED                           2588      3175      55        55

[mikel-brostrom]

I think this document is very vague, and I don't know exactly how to run the evaluation.

You clearly managed to run an evaluation. Regarding you negative MOTA, it does not mean that the evaluation is failing: #938

[wx-ikll007700]

I feel the same way. I think this document is very vague, and I don't know exactly how to run the evaluation. Also, I am evaluating on my own training dataset.

All sequences for  finished in 0.72 seconds

HOTA: -pedestrian                  HOTA      DetA      AssA      DetRe     DetPr     AssRe     AssPr     LocA      OWTA      HOTA(0)   LocA(0)   HOTALocA(0)
mot_particle                       36.257    30.554    44.278    36.564    44.857    48.719    60.939    67.89     40.074    81.835    45.642    37.351
COMBINED                           36.257    30.554    44.278    36.564    44.857    48.719    60.939    67.89     40.074    81.835    45.642    37.351

CLEAR: -pedestrian                 MOTA      MOTP      MODA      CLR_Re    CLR_Pr    MTR       PTR       MLR       sMOTA     CLR_TP    CLR_FN    CLR_FP    IDSW      MT        PT        ML        Frag
mot_particle                       -18.079   67.756    -18.016   31.748    38.949    3.6364    54.545    41.818    -28.316   1008      2167      1580      2         2         30        23        153
COMBINED                           -18.079   67.756    -18.016   31.748    38.949    3.6364    54.545    41.818    -28.316   1008      2167      1580      2         2         30        23        153

Identity: -pedestrian              IDF1      IDR       IDP       IDTP      IDFN      IDFP
mot_particle                       34.669    31.465    38.601    999       2176      1589
COMBINED                           34.669    31.465    38.601    999       2176      1589

Count: -pedestrian                 Dets      GT_Dets   IDs       GT_IDs
mot_particle                       2588      3175      55        55
COMBINED                           2588      3175      55        55

I feel that your evaluation process is normal, but the accuracy is low. It may be that the text you used for evaluation is not set properly.