result

[zhanghongruiupup]

hi,I use this loss for REID,but map result not good. while I think <n_mask = 1- p_mask> make error,should<n_mask = ~p_mask>

2019-11-14 16:11:22,232 reid_baseline.train INFO: Epoch[1] Iteration[20/183] Loss: 0.406, Acc: 0.001, Base Lr: 3.27e-05
2019-11-14 16:11:27,567 reid_baseline.train INFO: Epoch[1] Iteration[40/183] Loss: 0.340, Acc: 0.000, Base Lr: 3.27e-05
2019-11-14 16:11:33,046 reid_baseline.train INFO: Epoch[1] Iteration[60/183] Loss: 0.286, Acc: 0.000, Base Lr: 3.27e-05
2019-11-14 16:11:38,594 reid_baseline.train INFO: Epoch[1] Iteration[80/183] Loss: 0.246, Acc: 0.000, Base Lr: 3.27e-05
2019-11-14 16:11:44,225 reid_baseline.train INFO: Epoch[1] Iteration[100/183] Loss: 0.215, Acc: 0.001, Base Lr: 3.27e-05
2019-11-14 16:11:49,750 reid_baseline.train INFO: Epoch[1] Iteration[120/183] Loss: 0.192, Acc: 0.001, Base Lr: 3.27e-05
2019-11-14 16:11:55,047 reid_baseline.train INFO: Epoch[1] Iteration[140/183] Loss: 0.173, Acc: 0.002, Base Lr: 3.27e-05
2019-11-14 16:12:00,430 reid_baseline.train INFO: Epoch[1] Iteration[160/183] Loss: 0.160, Acc: 0.002, Base Lr: 3.27e-05
2019-11-14 16:12:05,996 reid_baseline.train INFO: Epoch[1] Iteration[180/183] Loss: 0.150, Acc: 0.001, Base Lr: 3.27e-05
2019-11-14 16:12:06,877 reid_baseline.train INFO: Epoch 1 done. Time per batch: 52.567[s] Speed: 222.8[samples/s]
2019-11-14 16:12:06,877 reid_baseline.train INFO: ----------
2019-11-14 16:12:12,795 reid_baseline.train INFO: Epoch[2] Iteration[20/184] Loss: 0.112, Acc: 0.000, Base Lr: 6.24e-05
2019-11-14 16:12:18,348 reid_baseline.train INFO: Epoch[2] Iteration[40/184] Loss: 0.113, Acc: 0.000, Base Lr: 6.24e-05
2019-11-14 16:12:23,765 reid_baseline.train INFO: Epoch[2] Iteration[60/184] Loss: 0.109, Acc: 0.001, Base Lr: 6.24e-05
2019-11-14 16:12:29,121 reid_baseline.train INFO: Epoch[2] Iteration[80/184] Loss: 0.107, Acc: 0.001, Base Lr: 6.24e-05
2019-11-14 16:12:34,911 reid_baseline.train INFO: Epoch[2] Iteration[100/184] Loss: 0.104, Acc: 0.000, Base Lr: 6.24e-05
2019-11-14 16:12:40,585 reid_baseline.train INFO: Epoch[2] Iteration[120/184] Loss: 0.101, Acc: 0.001, Base Lr: 6.24e-05
2019-11-14 16:12:46,198 reid_baseline.train INFO: Epoch[2] Iteration[140/184] Loss: 0.098, Acc: 0.002, Base Lr: 6.24e-05
2019-11-14 16:12:51,756 reid_baseline.train INFO: Epoch[2] Iteration[160/184] Loss: 0.096, Acc: 0.002, Base Lr: 6.24e-05
2019-11-14 16:12:57,291 reid_baseline.train INFO: Epoch[2] Iteration[180/184] Loss: 0.093, Acc: 0.003, Base Lr: 6.24e-05
2019-11-14 16:12:58,450 reid_baseline.train INFO: Epoch 2 done. Time per batch: 50.719[s] Speed: 232.2[samples/s]
2019-11-14 16:12:58,451 reid_baseline.train INFO: ----------
2019-11-14 16:13:04,366 reid_baseline.train INFO: Epoch[3] Iteration[20/183] Loss: 0.092, Acc: 0.001, Base Lr: 9.21e-05
2019-11-14 16:13:09,908 reid_baseline.train INFO: Epoch[3] Iteration[40/183] Loss: 0.090, Acc: 0.001, Base Lr: 9.21e-05
2019-11-14 16:13:15,367 reid_baseline.train INFO: Epoch[3] Iteration[60/183] Loss: 0.088, Acc: 0.002, Base Lr: 9.21e-05
2019-11-14 16:13:20,672 reid_baseline.train INFO: Epoch[3] Iteration[80/183] Loss: 0.086, Acc: 0.002, Base Lr: 9.21e-05
2019-11-14 16:13:26,072 reid_baseline.train INFO: Epoch[3] Iteration[100/183] Loss: 0.085, Acc: 0.002, Base Lr: 9.21e-05
2019-11-14 16:13:31,729 reid_baseline.train INFO: Epoch[3] Iteration[120/183] Loss: 0.083, Acc: 0.002, Base Lr: 9.21e-05

this my train result, Acc not good