Hi, first of all thank you for your excellent work.
I encountered some problems when trying to train, when running train_cstrack.py, training crowdhuman, then only GIou Loss, id loss is 0. Then refer to the gen_crowd_id.py code in FairMOT to generate new labels. During the training process, the size of the weight model is very large, and the loss is also unstable. What is the reason?
Hi, first of all thank you for your excellent work. I encountered some problems when trying to train, when running train_cstrack.py, training crowdhuman, then only GIou Loss, id loss is 0. Then refer to the gen_crowd_id.py code in FairMOT to generate new labels. During the training process, the size of the weight model is very large, and the loss is also unstable. What is the reason?