Question about MOT17 dataset

[ckcraig01]

Dear Author:

As you have mentioned, you used full mot17 for training the mot17 mix_det

But I found two issues (1) https://github.com/ifzhang/ByteTrack/blob/57d86bd465a09494ca1e8de7d013857e7deaddf4/tools/convert_mot17_to_coco.py#L38

if split = 'train', and not FRCNN in seq, then the folder would not be included inside training set, (the question is: why SDP and DPM are not used for training)

(2) https://github.com/ifzhang/ByteTrack/blob/57d86bd465a09494ca1e8de7d013857e7deaddf4/tools/mix_data_test_mot17.py#L19

only use half train for mix_det (not consistent as you mentioned using full mot17.)

Thanks for your answer

[ckcraig01]

Let me answer the question (1) by my self, because FRCNN/SDP/DPM are from different detectors, so choose one set would be enough But I am still curious about question (2)

[YamacTan]

Hi @ckcraig01 , Did you find a way to get the results of different detectors?

I noticed the error on the same line in MOT17 conversion script and changed it. But the track.py script still only produces results via FRCNN. I couldn't find any args to specify the detector either.

Bests, Y.

Edit: I think the situation is probably due to COCO API. But still no solution found.