How do you train

[myh12138]

In your code , i do not find a code to make val dataset , but still need val set.

[timmeinhardt]

There are no validation sets for the MOTChallenge datasets. You need to do cross-validation and then submit your best results to the test server.

[myh12138]

There are no validation sets for the MOTChallenge datasets. You need to do cross-validation and then submit your best results to the test server.

Can you share your training parameters?

[timmeinhardt]

We train with the default parameters in the Faster R-CNN training notebook (experiments/scripts/faster_rcnn_fpn_training.ipynb). A cross-validation can be done by chaning the test_split_seqs variable in the notebook.