facebookresearch / detr

End-to-End Object Detection with Transformers
Apache License 2.0
13.08k stars 2.37k forks source link

What does the different loss curves while training mean? #594

Open Ashwin-Dhakal opened 11 months ago

Ashwin-Dhakal commented 11 months ago

❓What are the meaning of all these different train and test loss types while training the model?

{"train_lr": 9.999999999999975e-05, "train_class_error": -4.6238754734848485e-08, "train_loss": 8.158622120365951, "train_loss_ce": 7.72964473471251e-06, "train_loss_bbox": 0.1408537747959296, "train_loss_giou": 1.2125587486859524, "train_loss_ce_0": 1.8782344399038595e-05, "train_loss_bbox_0": 0.1459837333496773, "train_loss_giou_0": 1.2347586142294336, "train_loss_ce_1": 1.54084060602944e-05, "train_loss_bbox_1": 0.14215930871891252, "train_loss_giou_1": 1.2195609208309288, "train_loss_ce_2": 1.4771415453651571e-05, "train_loss_bbox_2": 0.14131249537070592, "train_loss_giou_2": 1.214312648411953, "train_loss_ce_3": 1.21140988622673e-05, "train_loss_bbox_3": 0.14104107266122645, "train_loss_giou_3": 1.2126423420328083, "train_loss_ce_4": 8.637233664751075e-06, "train_loss_bbox_4": 0.14088943052020939, "train_loss_giou_4": 1.2124716023604074, "train_loss_ce_unscaled": 7.72964473471251e-06, "train_class_error_unscaled": -4.6238754734848485e-08, "train_loss_bbox_unscaled": 0.0281707549693458, "train_loss_giou_unscaled": 0.6062793743429762, "train_cardinality_error_unscaled": 271.2590909090909, "train_loss_ce_0_unscaled": 1.8782344399038595e-05, "train_loss_bbox_0_unscaled": 0.029196746748956767, "train_loss_giou_0_unscaled": 0.6173793071147168, "train_cardinality_error_0_unscaled": 271.2590909090909, "train_loss_ce_1_unscaled": 1.54084060602944e-05, "train_loss_bbox_1_unscaled": 0.02843186174039588, "train_loss_giou_1_unscaled": 0.6097804604154644, "train_cardinality_error_1_unscaled": 271.2590909090909, "train_loss_ce_2_unscaled": 1.4771415453651571e-05, "train_loss_bbox_2_unscaled": 0.02826249905382142, "train_loss_giou_2_unscaled": 0.6071563242059765, "train_cardinality_error_2_unscaled": 271.2590909090909, "train_loss_ce_3_unscaled": 1.21140988622673e-05, "train_loss_bbox_3_unscaled": 0.028208214580786952, "train_loss_giou_3_unscaled": 0.6063211710164041, "train_cardinality_error_3_unscaled": 271.2590909090909, "train_loss_ce_4_unscaled": 8.637233664751075e-06, "train_loss_bbox_4_unscaled": 0.0281778860611446, "train_loss_giou_4_unscaled": 0.6062358011802037, "train_cardinality_error_4_unscaled": 271.2590909090909, "test_class_error": 2.2439395680147059e-07, "test_loss": 7.7116100788116455, "test_loss_ce": 4.31399909342979e-06, "test_loss_bbox": 0.11455829638768644, "test_loss_giou": 1.160562935997458, "test_loss_ce_0": 9.774426563821822e-06, "test_loss_bbox_0": 0.11711135265581749, "test_loss_giou_0": 1.2095311056165134, "test_loss_ce_1": 7.848421098381496e-06, "test_loss_bbox_1": 0.11288979868678485, "test_loss_giou_1": 1.1748274617335375, "test_loss_ce_2": 8.424342956245442e-06, "test_loss_bbox_2": 0.11273422661949606, "test_loss_giou_2": 1.161564537707497, "test_loss_ce_3": 7.18141352669644e-06, "test_loss_bbox_3": 0.1133256427505437, "test_loss_giou_3": 1.1599872988813065, "test_loss_ce_4": 4.76531905700112e-06, "test_loss_bbox_4": 0.11427368333234507, "test_loss_giou_4": 1.1602014180491953, "test_loss_ce_unscaled": 4.31399909342979e-06, "test_class_error_unscaled": 2.2439395680147059e-07, "test_loss_bbox_unscaled": 0.022911659167969927, "test_loss_giou_unscaled": 0.580281467998729, "test_cardinality_error_unscaled": 323.51213253245635, "test_loss_ce_0_unscaled": 9.774426563821822e-06, "test_loss_bbox_0_unscaled": 0.02342227062977412, "test_loss_giou_0_unscaled": 0.6047655528082567, "test_cardinality_error_0_unscaled": 323.51213253245635, "test_loss_ce_1_unscaled": 7.848421098381496e-06, "test_loss_bbox_1_unscaled": 0.02257795974283534, "test_loss_giou_1_unscaled": 0.5874137308667687, "test_cardinality_error_1_unscaled": 323.51213253245635, "test_loss_ce_2_unscaled": 8.424342956245442e-06, "test_loss_bbox_2_unscaled": 0.022546845241723693, "test_loss_giou_2_unscaled": 0.5807822688537485, "test_cardinality_error_2_unscaled": 323.51213253245635, "test_loss_ce_3_unscaled": 7.18141352669644e-06, "test_loss_bbox_3_unscaled": 0.0226651285227169, "test_loss_giou_3_unscaled": 0.5799936494406532, "test_cardinality_error_3_unscaled": 323.51213253245635, "test_loss_ce_4_unscaled": 4.76531905700112e-06, "test_loss_bbox_4_unscaled": 0.022854736677425748, "test_loss_giou_4_unscaled": 0.5801007090245976, "test_cardinality_error_4_unscaled": 323.51213253245635, "test_coco_eval_bbox": [0.0, 0.0, 0.0, 0.0, -1.0, -1.0, 0.0, 0.0, 0.0, 0.0, -1.0, -1.0], "epoch": 18, "n_parameters": 41381638}

Ashwin-Dhakal commented 11 months ago

Here are the curves: image