mask_loss and iou_score_loss is always 0

[aguang1201]

Thank you for your work, mask_loss and iou_score_loss always 0 in the process of training my own data, what could be the reason? 2022-07-05 19:20:48,783 - INFO - Training 2022-07-05 19:21:01,139 - INFO - Epoch [1/40][1/660] lr: 0.003, eta: 3 days, 18:35:15, mem: 6372, data_time: 4.74, iter_time: 12.35, semantic_loss: 0.3645, offset_loss: 0.8161, cls_loss: 1.1041, mask_loss: 0.0000, iou_score_loss: 0.0000, loss: 2.2847 2022-07-05 19:21:06,520 - INFO - Epoch [1/40][2/660] lr: 0.003, eta: 2 days, 16:59:03, mem: 6372, data_time: 0.00, iter_time: 5.37, semantic_loss: 0.6133, offset_loss: 1.2717, cls_loss: 0.9968, mask_loss: 0.0000, iou_score_loss: 0.0000, loss: 2.8818 2022-07-05 19:21:08,048 - INFO - Epoch [1/40][3/660] lr: 0.003, eta: 1 day, 23:04:01, mem: 6372, data_time: 0.01, iter_time: 1.53, semantic_loss: 0.2472, offset_loss: 0.8631, cls_loss: 1.0198, mask_loss: 0.0000, iou_score_loss: 0.0000, loss: 2.1302 2022-07-05 19:21:09,213 - INFO - Epoch [1/40][4/660] lr: 0.003, eta: 1 day, 13:26:44, mem: 6372, data_time: 0.01, iter_time: 1.17, semantic_loss: 0.2269, offset_loss: 0.4879, cls_loss: 0.9860, mask_loss: 0.0000, iou_score_loss: 0.0000, loss: 1.7008 2022-07-05 19:21:13,257 - INFO - Epoch [1/40][5/660] lr: 0.003, eta: 1 day, 11:53:11, mem: 6372, data_time: 0.00, iter_time: 4.04, semantic_loss: 0.3312, offset_loss: 0.2553, cls_loss: 0.9681, mask_loss: 0.0000, iou_score_loss: 0.0000, loss: 1.5546

[thangvubk]

Mask loss and iou_score_loss are on postitive samples only which have IoU overlap higher than 0.5. May be your clustering radius is not appropriate leading to no positive samples.