search

ifzhang / ByteTrack

[ECCV 2022] ByteTrack: Multi-Object Tracking by Associating Every Detection Box
MIT License
4.69k stars 891 forks source link

The bbox of the detection results obtained on the MOT17 dataset has negative values #175

Open zyiyid opened 2 years ago

zyiyid commented 2 years ago

image The bbox of the detection results obtained on the MOT17 dataset has negative values.What is the reason for this and how can I adjust it?

zyiyid commented 2 years ago

I want to get all detection results of MOT17 training set. I use yolox_x_mix_det.py and change its val_ann to 'train.json' and the MOTDataset name to 'train'. But there are negative values in the TXT file I get after running tools/track.py. How can I get the same format as the official MOT Challenge detection result?

LamnouarMohamed commented 2 years ago

Hi

sorry for deleting my previous answer because my previous answer is for another question like yours, I think problem come from when you resize image. (when I was working with this repository I meet with problem a little bit like your problem but my problem is the value of detection (x, y, w, h) is higher than it's dimension of image, the solution of the problem just take care when you resize your image.

zyiyid commented 2 years ago

Hi

sorry for deleting my previous answer because my previous answer is for another question like yours, I think problem come from when you resize image. (when I was working with this repository I meet with problem a little bit like your problem but my problem is the value of detection (x, y, w, h) is higher than it's dimension of image, the solution of the problem just take care when you resize your image.

Hi, how did you end up doing this? As far as I know, the input size of yolox_x_mix_det.py is 800x1440, the original input size of yolox_x is 640x640, and the image size of MOT17 set is 1920x1080. I don't know why it is set this way and how can I modify it?