VisDrone dataset

[nguyenpbui]

Dear PeizeSun,
First of all, I want to say thank you for your hard work and this model is very awesome.
I am applying your work to VisDrone dataset which is very similar to MOT dataset. I wrote some new files to run training with VisDrone.
I trained it for the first time with 20 epochs and I got very bad result (in the image attached). In the original training .sh file, you trained your model with 150 epochs but I just ran with 20 epochs. Does it affect the results?
Thank you.

[PeizeSun]

Hi~
150epoch is definitely better than 20epoch.
However, you only get 1.6 mAP after 20epoch, it seems something wrong happens.

[nguyenpbui]

Thank you for your quick reply.
I totally agree that 150 epochs is much better than 20. With 150 epochs, you achieve 74.5 accuracy on MOT17, so I wonder when I train with only 20 epoch, can I get 10 or 20 accuracy instead of 1.6 in the image attached?