IOU loss is always 0

[hxue3]

I am very new to ByteTrack. I tried to train a model using my own dataset. I convert my dataset into mot17 data format first then using convert_mot17_coco.py to convert to coco format.

However, I noticed the training process shows iou loss is always 0.

2022-04-05 01:09:52 | INFO | yolox.core.trainer:242 - epoch: 2/10, iter: 20/530, mem: 26620Mb, iter_time: 0.558s, data_time: 0.001s, total_loss: 0.032, iou_loss: 0.000, l1_loss: 0.000, conf_loss: 0.032, cls_loss: 0.000, lr: 3.750e-05, size: 832, ETA: 0:57:10 2022-04-05 01:10:04 | INFO | yolox.core.trainer:242 - epoch: 2/10, iter: 40/530, mem: 26620Mb, iter_time: 0.627s, data_time: 0.001s, total_loss: 0.018, iou_loss: 0.000, l1_loss: 0.000, conf_loss: 0.018, cls_loss: 0.000, lr: 3.750e-05, size: 576, ETA: 0:56:39 2022-04-05 01:10:12 | INFO | yolox.core.trainer:242 - epoch: 2/10, iter: 60/530, mem: 26620Mb, iter_time: 0.382s, data_time: 0.001s, total_loss: 0.028, iou_loss: 0.000, l1_loss: 0.000, conf_loss: 0.028, cls_loss: 0.000, lr: 3.750e-05, size: 768, ETA: 0:55:31 2022-04-05 01:10:26 | INFO | yolox.core.trainer:242 - epoch: 2/10, iter: 80/530, mem: 26620Mb, iter_time: 0.688s, data_time: 0.001s, total_loss: 0.017, iou_loss: 0.000, l1_loss: 0.000, conf_loss: 0.017, cls_loss: 0.000, lr: 3.750e-05, size: 576, ETA: 0:55:14 2022-04-05 01:10:37 | INFO | yolox.core.trainer:242 - epoch: 2/10, iter: 100/530, mem: 26620Mb, iter_time: 0.547s, data_time: 0.001s, total_loss: 0.026, iou_loss: 0.000, l1_loss: 0.000, conf_loss: 0.026, cls_loss: 0.000, lr: 3.750e-05, size:768, ETA: 0:54:36 2022-04-05 01:10:46 | INFO | yolox.core.trainer:242 - epoch: 2/10, iter: 120/530, mem: 26620Mb, iter_time: 0.478s, data_time: 0.001s, total_loss: 0.029, iou_loss: 0.000, l1_loss: 0.000, conf_loss: 0.029, cls_loss: 0.000, lr: 3.750e-05, size:800, ETA: 0:53:50 2022-04-05 01:11:00 | INFO | yolox.core.trainer:242 - epoch: 2/10, iter: 140/530, mem: 26620Mb, iter_time: 0.669s, data_time: 0.002s, total_loss: 0.051, iou_loss: 0.000, l1_loss: 0.000, conf_loss: 0.051, cls_loss: 0.000, lr: 3.750e-05, size:1024, ETA: 0:53:33 2022-04-05 01:11:17 | INFO | yolox.core.trainer:242 - epoch: 2/10, iter: 160/530, mem: 26620Mb, iter_time: 0.840s, data_time: 0.001s, total_loss: 0.042, iou_loss: 0.000, l1_loss: 0.000, conf_loss: 0.042, cls_loss: 0.000, lr: 3.750e-05, size:992, ETA: 0:53:38 2022-04-05 01:11:29 | INFO | yolox.core.trainer:242 - epoch: 2/10, iter: 180/530, mem: 26620Mb, iter_time: 0.610s, data_time: 0.001s, total_loss: 0.023, iou_loss: 0.000, l1_loss: 0.000, conf_loss: 0.023, cls_loss: 0.000, lr: 3.750e-05, size:704, ETA: 0:53:13 2022-04-05 01:11:38 | INFO | yolox.core.trainer:242 - epoch: 2/10, iter: 200/530, mem: 26620Mb, iter_time: 0.455s, data_time: 0.001s, total_loss: 0.020, iou_loss: 0.000, l1_loss: 0.000, conf_loss: 0.020, cls_loss: 0.000, lr: 3.750e-05, size:640, ETA: 0:52:29 2022-04-05 01:11:49 | INFO | yolox.core.trainer:242 - epoch: 2/10, iter: 220/530, mem: 26620Mb, iter_time: 0.538s, data_time: 0.002s, total_loss: 0.020, iou_loss: 0.000, l1_loss: 0.000, conf_loss: 0.020, cls_loss: 0.000, lr: 3.750e-05, size:672, ETA: 0:51:57 2022-04-05 01:12:01 | INFO | yolox.core.trainer:242 - epoch: 2/10, iter: 240/530, mem: 26620Mb, iter_time: 0.614s, data_time: 0.001s, total_loss: 0.016, iou_loss: 0.000, l1_loss: 0.000, conf_loss: 0.016, cls_loss: 0.000, lr: 3.750e-05, size:576, ETA: 0:51:35 2022-04-05 01:12:09 | INFO | yolox.core.trainer:242 - epoch: 2/10, iter: 260/530, mem: 26620Mb, iter_time: 0.398s, data_time: 0.001s, total_loss: 0.043, iou_loss: 0.000, l1_loss: 0.000, conf_loss: 0.043, cls_loss: 0.000, lr: 3.750e-05, size:992, ETA: 0:50:48 2022-04-05 01:12:23 | INFO | yolox.core.trainer:242 - epoch: 2/10, iter: 280/530, mem: 26620Mb, iter_time: 0.701s, data_time: 0.001s, total_loss: 0.046, iou_loss: 0.000, l1_loss: 0.000, conf_loss: 0.046, cls_loss: 0.000, lr: 3.750e-05, size:1024, ETA: 0:50:38

After finisihed training, I tried to use trained model to test on a video, but it cannot detect any objects.

Anyone can help me with this?