about IDF1 IDP IDR .

[kingwmk]

Hello, I ran the model you proposed, but the IDF1 IDP IDR results, which performance is very bad, are different from yours. What could be the cause? looking forward to your reply.
Thank you very much! Sincerely,

Evaluating ...
... MOT16-02
Preprocessing (cleaning) MOT16-02...
......
Removing 626 boxes from solution...
*** 2D (Bounding Box overlap) ***
IDF1 IDP IDR| Rcll Prcn FAR| GT MT PT ML| FP FN IDs FM| MOTA MOTP MOTAL
23.0 66.9 13.9| 30.7 90.3 0.98| 54 7 21 26| 587 12356 77 179| 27.0 75.2 27.4

... MOT16-05

Preprocessing (cleaning) MOT16-05...
........
Removing 1 boxes from solution...
*** 2D (Bounding Box overlap) ***
IDF1 IDP IDR| Rcll Prcn FAR| GT MT PT ML| FP FN IDs FM| MOTA MOTP MOTAL
34.7 56.7 25.0| 45.0 90.9 0.37| 125 16 68 41| 307 3747 75 140| 39.4 75.6 40.5

... MOT16-09

Preprocessing (cleaning) MOT16-09...
.....
Removing 758 boxes from solution...
*** 2D (Bounding Box overlap) ***
IDF1 IDP IDR| Rcll Prcn FAR| GT MT PT ML| FP FN IDs FM| MOTA MOTP MOTAL
38.4 66.1 27.1| 63.5 92.3 0.53| 25 9 13 3| 280 1921 57 77| 57.0 75.2 58.1

... MOT16-11

Preprocessing (cleaning) MOT16-11...
.........
Removing 2 boxes from solution...
*** 2D (Bounding Box overlap) ***
IDF1 IDP IDR| Rcll Prcn FAR| GT MT PT ML| FP FN IDs FM| MOTA MOTP MOTAL
51.1 71.0 39.9| 58.1 94.0 0.38| 69 11 29 29| 340 3844 42 78| 53.9 79.2 54.4

... MOT16-13

Preprocessing (cleaning) MOT16-13...
.......
Removing 0 boxes from solution...
*** 2D (Bounding Box overlap) ***
IDF1 IDP IDR| Rcll Prcn FAR| GT MT PT ML| FP FN IDs FM| MOTA MOTP MOTAL
22.9 62.0 14.1| 30.8 80.7 1.13| 107 11 40 56| 844 7922 60 198| 22.9 72.2 23.4

********************* Your Benchmark Results (2D) ***********************
IDF1 IDP IDR| Rcll Prcn FAR| GT MT PT ML| FP FN IDs FM| MOTA MOTP MOTAL
20.2 43.0 13.2| 41.0 89.8 0.65| 380 54 171 155| 2358 29790 311 672| 35.8 75.8 36.4

[longcw]

I am not sure what's your problem. Please check the new evaluation code in python.

[longcw]

https://github.com/longcw/MOTDT#evaluate

[kingwmk]

Thanks a lot!