Results cannot be reproduced？

[cq0907]

I train a model by:

After training 98 epoch, I get the following results:

Why my result are so different from yours?

The setting is:

[zesenwu23]

Hi, can you upload the training log for the first 10 epochs?

[cq0907]

Hi, can you upload the training log for the first 10 epochs?

Epoch: [2][100/695] Time: 0.272 (0.278) lr:0.030 Loss: 89.4270 (88.8914) iLoss: 1.4270 (0.8914) TLoss: 88.0000 (88.0000) KLoss: 8.2916 (8.5085) Accu: 38.63
Epoch: [2][150/695] Time: 0.275 (0.276) lr:0.030 Loss: 88.7448 (88.8946) iLoss: 0.7448 (0.8946) TLoss: 88.0000 (88.0000) KLoss: 8.2963 (8.5243) Accu: 38.67
Epoch: [2][200/695] Time: 0.273 (0.276) lr:0.030 Loss: 88.8029 (88.8883) iLoss: 0.8029 (0.8883) TLoss: 88.0000 (88.0000) KLoss: 8.3028 (8.6061) Accu: 38.67
Epoch: [2][250/695] Time: 0.273 (0.275) lr:0.030 Loss: 88.3966 (88.8919) iLoss: 0.3966 (0.8919) TLoss: 88.0000 (88.0000) KLoss: 7.6682 (8.7429) Accu: 38.68
Epoch: [2][300/695] Time: 0.273 (0.275) lr:0.030 Loss: 88.9312 (88.8849) iLoss: 0.9312 (0.8849) TLoss: 88.0000 (88.0000) KLoss: 8.1458 (8.8047) Accu: 38.83
Epoch: [2][350/695] Time: 0.273 (0.275) lr:0.030 Loss: 88.9140 (88.8733) iLoss: 0.9140 (0.8733) TLoss: 88.0000 (88.0000) KLoss: 8.9773 (8.8654) Accu: 38.94
Epoch: [2][400/695] Time: 0.273 (0.275) lr:0.030 Loss: 88.6361 (88.8473) iLoss: 0.6361 (0.8473) TLoss: 88.0000 (88.0000) KLoss: 9.0357 (8.8763) Accu: 39.25
Epoch: [2][450/695] Time: 0.282 (0.274) lr:0.030 Loss: 88.4686 (88.8283) iLoss: 0.4686 (0.8283) TLoss: 88.0000 (88.0000) KLoss: 9.5262 (8.8743) Accu: 39.52
Epoch: [2][500/695] Time: 0.273 (0.274) lr:0.030 Loss: 88.6134 (88.8055) iLoss: 0.6134 (0.8055) TLoss: 88.0000 (88.0000) KLoss: 9.1093 (8.8821) Accu: 39.80
Epoch: [2][550/695] Time: 0.273 (0.274) lr:0.030 Loss: 88.6441 (88.7872) iLoss: 0.6441 (0.7872) TLoss: 88.0000 (88.0000) KLoss: 9.5906 (8.8930) Accu: 39.99
Epoch: [2][600/695] Time: 0.272 (0.274) lr:0.030 Loss: 88.1605 (88.7717) iLoss: 0.1605 (0.7717) TLoss: 88.0000 (88.0000) KLoss: 8.2677 (8.9089) Accu: 40.20
Epoch: [2][650/695] Time: 0.273 (0.274) lr:0.030 Loss: 88.3207 (88.7560) iLoss: 0.3207 (0.7560) TLoss: 88.0000 (88.0000) KLoss: 8.3744 (8.9002) Accu: 40.40
Test Epoch: 2
Extracting Gallery Feature...
Extracting Time: 3.557
Extracting Query Feature...
Extracting Time: 5.575
Evaluation Time: 4.138
POOL: Rank-1: 16.51% | Rank-5: 36.94% | Rank-10: 48.36%| Rank-20: 62.79%| mAP: 15.83%| mINP: 7.45%
FC: Rank-1: 21.11% | Rank-5: 45.31% | Rank-10: 59.77%| Rank-20: 75.49%| mAP: 20.87%| mINP: 10.63%
Best Epoch [2]
==> Preparing Data Loader...
3
[18948 18933 18940 ... 17397 17443 17421]
[8563 8548 8563 ... 7494 7504 7498]
Epoch: [3][0/695] Time: 0.716 (0.716) lr:0.040 Loss: 88.6011 (88.6011) iLoss: 0.6011 (0.6011) TLoss: 88.0000 (88.0000) KLoss: 8.6152 (8.6152) Accu: 41.15
Epoch: [3][50/695] Time: 0.273 (0.282) lr:0.040 Loss: 88.3976 (88.6086) iLoss: 0.3976 (0.6086) TLoss: 88.0000 (88.0000) KLoss: 9.9551 (8.8652) Accu: 42.22
Epoch: [3][100/695] Time: 0.272 (0.278) lr:0.040 Loss: 88.9570 (88.6278) iLoss: 0.9570 (0.6278) TLoss: 88.0000 (88.0000) KLoss: 8.7417 (8.9135) Accu: 42.08
Epoch: [3][150/695] Time: 0.274 (0.276) lr:0.040 Loss: 88.3955 (88.6135) iLoss: 0.3955 (0.6135) TLoss: 88.0000 (88.0000) KLoss: 8.9246 (8.9067) Accu: 42.14
Epoch: [3][200/695] Time: 0.274 (0.276) lr:0.040 Loss: 88.6157 (88.6116) iLoss: 0.6157 (0.6116) TLoss: 88.0000 (88.0000) KLoss: 9.3078 (8.9218) Accu: 42.15
Epoch: [3][250/695] Time: 0.274 (0.275) lr:0.040 Loss: 88.8796 (88.6140) iLoss: 0.8796 (0.6140) TLoss: 88.0000 (88.0000) KLoss: 9.9581 (8.9296) Accu: 42.08
Epoch: [3][300/695] Time: 0.274 (0.275) lr:0.040 Loss: 88.3863 (88.6050) iLoss: 0.3863 (0.6050) TLoss: 88.0000 (88.0000) KLoss: 8.9333 (8.9124) Accu: 42.18
Epoch: [3][350/695] Time: 0.273 (0.275) lr:0.040 Loss: 88.2817 (88.5953) iLoss: 0.2817 (0.5953) TLoss: 88.0000 (88.0000) KLoss: 8.7242 (8.9181) Accu: 42.30
Epoch: [3][400/695] Time: 0.273 (0.275) lr:0.040 Loss: 88.3939 (88.5864) iLoss: 0.3939 (0.5864) TLoss: 88.0000 (88.0000) KLoss: 8.1597 (8.9067) Accu: 42.39
Epoch: [3][450/695] Time: 0.274 (0.275) lr:0.040 Loss: 88.2766 (88.5786) iLoss: 0.2766 (0.5786) TLoss: 88.0000 (88.0000) KLoss: 8.1062 (8.8791) Accu: 42.50
Epoch: [3][500/695] Time: 0.273 (0.275) lr:0.040 Loss: 89.1530 (88.5682) iLoss: 1.1530 (0.5682) TLoss: 88.0000 (88.0000) KLoss: 9.7190 (8.8280) Accu: 42.66
Epoch: [3][550/695] Time: 0.274 (0.274) lr:0.040 Loss: 88.3128 (88.5551) iLoss: 0.3128 (0.5551) TLoss: 88.0000 (88.0000) KLoss: 8.3632 (8.7564) Accu: 42.85
Epoch: [3][600/695] Time: 0.274 (0.274) lr:0.040 Loss: 88.4823 (88.5447) iLoss: 0.4823 (0.5447) TLoss: 88.0000 (88.0000) KLoss: 8.5399 (8.7063) Accu: 42.97
Epoch: [3][650/695] Time: 0.272 (0.274) lr:0.040 Loss: 88.4157 (88.5339) iLoss: 0.4157 (0.5339) TLoss: 88.0000 (88.0000) KLoss: 7.1020 (8.6463) Accu: 43.12
==> Preparing Data Loader...
4
[6088 6095 6073 ... 1956 1884 1946]
[2636 2611 2614 ... 774 774 791]
Epoch: [4][0/695] Time: 0.700 (0.700) lr:0.050 Loss: 88.5157 (88.5157) iLoss: 0.5157 (0.5157) TLoss: 88.0000 (88.0000) KLoss: 8.2325 (8.2325) Accu: 43.23
Epoch: [4][50/695] Time: 0.273 (0.282) lr:0.050 Loss: 88.4549 (88.4709) iLoss: 0.4549 (0.4709) TLoss: 88.0000 (88.0000) KLoss: 7.5466 (7.9749) Accu: 43.85
Epoch: [4][100/695] Time: 0.273 (0.278) lr:0.050 Loss: 88.3378 (88.4647) iLoss: 0.3378 (0.4647) TLoss: 88.0000 (88.0000) KLoss: 8.2959 (8.0032) Accu: 44.03
Epoch: [4][150/695] Time: 0.273 (0.276) lr:0.050 Loss: 88.4724 (88.4954) iLoss: 0.4724 (0.4954) TLoss: 88.0000 (88.0000) KLoss: 7.4439 (8.0202) Accu: 43.67
Epoch: [4][200/695] Time: 0.273 (0.276) lr:0.050 Loss: 88.4874 (88.4942) iLoss: 0.4874 (0.4942) TLoss: 88.0000 (88.0000) KLoss: 7.8594 (8.0040) Accu: 43.64
Epoch: [4][250/695] Time: 0.273 (0.275) lr:0.050 Loss: 88.7037 (88.4915) iLoss: 0.7037 (0.4915) TLoss: 88.0000 (88.0000) KLoss: 8.2477 (8.0059) Accu: 43.69
Epoch: [4][300/695] Time: 0.273 (0.275) lr:0.050 Loss: 88.6629 (88.4894) iLoss: 0.6629 (0.4894) TLoss: 88.0000 (88.0000) KLoss: 7.9568 (7.9894) Accu: 43.69
Epoch: [4][350/695] Time: 0.274 (0.275) lr:0.050 Loss: 88.3515 (88.4847) iLoss: 0.3515 (0.4847) TLoss: 88.0000 (88.0000) KLoss: 8.9861 (7.9794) Accu: 43.75
Epoch: [4][400/695] Time: 0.272 (0.275) lr:0.050 Loss: 88.5411 (88.4831) iLoss: 0.5411 (0.4834) TLoss: 88.0000 (87.9997) KLoss: 8.5402 (7.9610) Accu: 43.75
Epoch: [4][450/695] Time: 0.273 (0.274) lr:0.050 Loss: 88.2826 (88.4732) iLoss: 0.2826 (0.4735) TLoss: 88.0000 (87.9997) KLoss: 7.8360 (7.9248) Accu: 43.86
Epoch: [4][500/695] Time: 0.273 (0.274) lr:0.050 Loss: 88.2636 (88.4644) iLoss: 0.2636 (0.4646) TLoss: 88.0000 (87.9998) KLoss: 6.3330 (7.8879) Accu: 43.96
Epoch: [4][550/695] Time: 0.272 (0.274) lr:0.050 Loss: 88.1958 (88.4555) iLoss: 0.1958 (0.4557) TLoss: 88.0000 (87.9998) KLoss: 7.0777 (7.8301) Accu: 44.08
Epoch: [4][600/695] Time: 0.272 (0.274) lr:0.050 Loss: 88.3473 (88.4462) iLoss: 0.3473 (0.4464) TLoss: 88.0000 (87.9998) KLoss: 6.7138 (7.7825) Accu: 44.18
Epoch: [4][650/695] Time: 0.272 (0.274) lr:0.050 Loss: 88.4024 (88.4374) iLoss: 0.4024 (0.4376) TLoss: 88.0000 (87.9998) KLoss: 6.4832 (7.7257) Accu: 44.31
Test Epoch: 4
Extracting Gallery Feature...
Extracting Time: 0.779
Extracting Query Feature...
Extracting Time: 4.350
Evaluation Time: 4.090
POOL: Rank-1: 19.85% | Rank-5: 41.13% | Rank-10: 53.17%| Rank-20: 67.84%| mAP: 18.52%| mINP: 8.58%
FC: Rank-1: 26.16% | Rank-5: 54.25% | Rank-10: 68.26%| Rank-20: 81.72%| mAP: 25.37%| mINP: 13.49%
Best Epoch [4]
==> Preparing Data Loader...
5
[18073 18002 18005 ... 3811 3863 3861]
[7818 7808 7819 ... 1657 1665 1673]
Epoch: [5][0/695] Time: 0.713 (0.713) lr:0.060 Loss: 88.1604 (88.1604) iLoss: 0.1604 (0.1604) TLoss: 88.0000 (88.0000) KLoss: 6.1973 (6.1973) Accu: 47.40
Epoch: [5][50/695] Time: 0.273 (0.281) lr:0.060 Loss: 88.3838 (88.3612) iLoss: 0.3838 (0.3612) TLoss: 88.0000 (88.0000) KLoss: 6.4952 (6.8451) Accu: 45.37
Epoch: [5][100/695] Time: 0.273 (0.277) lr:0.060 Loss: 88.3777 (88.3803) iLoss: 0.3777 (0.3803) TLoss: 88.0000 (88.0000) KLoss: 7.1354 (6.8482) Accu: 45.18
Epoch: [5][150/695] Time: 0.273 (0.276) lr:0.060 Loss: 88.2980 (88.3776) iLoss: 0.2980 (0.3776) TLoss: 88.0000 (88.0000) KLoss: 6.5694 (6.8887) Accu: 45.23
Epoch: [5][200/695] Time: 0.272 (0.275) lr:0.060 Loss: 88.4809 (88.3864) iLoss: 0.4809 (0.3864) TLoss: 88.0000 (88.0000) KLoss: 6.9022 (6.8921) Accu: 45.09
Epoch: [5][250/695] Time: 0.273 (0.275) lr:0.060 Loss: 88.2699 (88.3864) iLoss: 0.2699 (0.3864) TLoss: 88.0000 (88.0000) KLoss: 6.3708 (6.8662) Accu: 45.12
Epoch: [5][300/695] Time: 0.277 (0.275) lr:0.060 Loss: 88.3163 (88.3776) iLoss: 0.3163 (0.3776) TLoss: 88.0000 (88.0000) KLoss: 6.2582 (6.8536) Accu: 45.22
Epoch: [5][350/695] Time: 0.272 (0.275) lr:0.060 Loss: 88.1860 (88.3741) iLoss: 0.1860 (0.3741) TLoss: 88.0000 (88.0000) KLoss: 5.9200 (6.8137) Accu: 45.29
Epoch: [5][400/695] Time: 0.273 (0.274) lr:0.060 Loss: 88.1951 (88.3686) iLoss: 0.1951 (0.3686) TLoss: 88.0000 (88.0000) KLoss: 6.1012 (6.7744) Accu: 45.37
Epoch: [5][450/695] Time: 0.274 (0.274) lr:0.060 Loss: 88.5585 (88.3654) iLoss: 0.5585 (0.3654) TLoss: 88.0000 (88.0000) KLoss: 6.5597 (6.7349) Accu: 45.40
Epoch: [5][500/695] Time: 0.274 (0.274) lr:0.060 Loss: 88.4268 (88.3627) iLoss: 0.4268 (0.3627) TLoss: 88.0000 (88.0000) KLoss: 6.4857 (6.6911) Accu: 45.43
Epoch: [5][550/695] Time: 0.274 (0.274) lr:0.060 Loss: 88.0958 (88.3602) iLoss: 0.0958 (0.3602) TLoss: 88.0000 (88.0000) KLoss: 5.7822 (6.6420) Accu: 45.48
Epoch: [5][600/695] Time: 0.274 (0.274) lr:0.060 Loss: 21.2410 (87.7977) iLoss: 6.9059 (0.3980) TLoss: 14.3351 (87.3997) KLoss: 4.4297 (6.5908) Accu: 45.21
Epoch: [5][650/695] Time: 0.273 (0.274) lr:0.060 Loss: 7.0363 (81.7301) iLoss: 6.2938 (0.8793) TLoss: 0.7425 (80.8508) KLoss: 1.6672 (6.2382) Accu: 41.97
==> Preparing Data Loader...
6
[16271 16340 16342 ... 20425 20415 20423]
[ 7014 7047 7021 ... 10039 10044 10041]
Epoch: [6][0/695] Time: 0.686 (0.686) lr:0.070 Loss: 6.7556 (6.7556) iLoss: 6.0077 (6.0077) TLoss: 0.7479 (0.7479) KLoss: 1.2174 (1.2174) Accu: 0.52
Epoch: [6][50/695] Time: 0.277 (0.282) lr:0.070 Loss: 5.8677 (6.4769) iLoss: 5.1514 (5.7499) TLoss: 0.7163 (0.7270) KLoss: 1.3028 (1.2581) Accu: 1.79
Epoch: [6][100/695] Time: 0.273 (0.278) lr:0.070 Loss: 6.1910 (6.3815) iLoss: 5.4860 (5.6604) TLoss: 0.7050 (0.7212) KLoss: 1.3270 (1.2567) Accu: 2.03
Epoch: [6][150/695] Time: 0.282 (0.276) lr:0.070 Loss: 6.5671 (6.2769) iLoss: 5.8560 (5.5607) TLoss: 0.7111 (0.7163) KLoss: 1.3876 (1.2966) Accu: 2.74
Epoch: [6][200/695] Time: 0.272 (0.276) lr:0.070 Loss: 6.2862 (6.1988) iLoss: 5.5866 (5.4854) TLoss: 0.6996 (0.7134) KLoss: 1.6873 (1.3396) Accu: 3.32
Epoch: [6][250/695] Time: 0.278 (0.275) lr:0.070 Loss: 5.9715 (6.1229) iLoss: 5.2693 (5.4122) TLoss: 0.7022 (0.7108) KLoss: 1.5540 (1.3893) Accu: 4.17
Epoch: [6][300/695] Time: 0.273 (0.275) lr:0.070 Loss: 6.2065 (6.0643) iLoss: 5.5080 (5.3556) TLoss: 0.6984 (0.7087) KLoss: 1.5774 (1.4518) Accu: 5.44
Epoch: [6][350/695] Time: 0.272 (0.275) lr:0.070 Loss: 5.7395 (6.0074) iLoss: 5.0407 (5.3004) TLoss: 0.6989 (0.7070) KLoss: 2.0679 (1.5042) Accu: 6.57
Epoch: [6][400/695] Time: 0.273 (0.274) lr:0.070 Loss: 5.6376 (5.9538) iLoss: 4.9428 (5.2479) TLoss: 0.6948 (0.7059) KLoss: 1.8836 (1.5573) Accu: 7.01
Epoch: [6][450/695] Time: 0.272 (0.274) lr:0.070 Loss: 5.3541 (5.9083) iLoss: 4.6586 (5.2034) TLoss: 0.6954 (0.7048) KLoss: 1.9755 (1.6113) Accu: 7.83
Epoch: [6][500/695] Time: 0.273 (0.274) lr:0.070 Loss: 5.9325 (5.8653) iLoss: 5.2368 (5.1614) TLoss: 0.6957 (0.7039) KLoss: 2.3936 (1.6686) Accu: 8.88
Epoch: [6][550/695] Time: 0.273 (0.274) lr:0.070 Loss: 4.9913 (5.8203) iLoss: 4.2986 (5.1171) TLoss: 0.6927 (0.7031) KLoss: 2.3705 (1.7250) Accu: 9.90
Epoch: [6][600/695] Time: 0.273 (0.274) lr:0.070 Loss: 4.6970 (5.7804) iLoss: 4.0035 (5.0781) TLoss: 0.6935 (0.7024) KLoss: 2.2517 (1.7815) Accu: 11.08
Epoch: [6][650/695] Time: 0.273 (0.274) lr:0.070 Loss: 5.2744 (5.7394) iLoss: 4.5821 (5.0378) TLoss: 0.6923 (0.7017) KLoss: 3.0256 (1.8371) Accu: 12.47
Test Epoch: 6
Extracting Gallery Feature...
Extracting Time: 0.801
Extracting Query Feature...
Extracting Time: 4.326
Evaluation Time: 4.128
POOL: Rank-1: 1.37% | Rank-5: 7.07% | Rank-10: 13.91%| Rank-20: 26.35%| mAP: 3.65%| mINP: 2.11%
FC: Rank-1: 1.05% | Rank-5: 6.68% | Rank-10: 13.96%| Rank-20: 24.93%| mAP: 3.69%| mINP: 2.33%
Best Epoch [4]
==> Preparing Data Loader...
7
[19199 19201 19203 ... 16944 17016 16951]
[8819 8817 8822 ... 7317 7313 7313]
Epoch: [7][0/695] Time: 0.716 (0.716) lr:0.080 Loss: 5.4215 (5.4215) iLoss: 4.7287 (4.7287) TLoss: 0.6928 (0.6928) KLoss: 3.0571 (3.0571) Accu: 30.21
Epoch: [7][50/695] Time: 0.283 (0.282) lr:0.080 Loss: 5.4436 (5.1843) iLoss: 4.7481 (4.4912) TLoss: 0.6955 (0.6932) KLoss: 2.5890 (2.7173) Accu: 28.96
Epoch: [7][100/695] Time: 0.273 (0.278) lr:0.080 Loss: 5.1101 (5.1418) iLoss: 4.4171 (4.4488) TLoss: 0.6930 (0.6930) KLoss: 2.6546 (2.7395) Accu: 29.82
Epoch: [7][150/695] Time: 0.271 (0.276) lr:0.080 Loss: 4.7229 (5.0963) iLoss: 4.0310 (4.4034) TLoss: 0.6919 (0.6929) KLoss: 2.7361 (2.7750) Accu: 30.57
Epoch: [7][200/695] Time: 0.274 (0.275) lr:0.080 Loss: 4.6991 (5.0435) iLoss: 4.0098 (4.3512) TLoss: 0.6893 (0.6923) KLoss: 3.1200 (2.8529) Accu: 32.34
Epoch: [7][250/695] Time: 0.273 (0.275) lr:0.080 Loss: 5.7062 (5.0283) iLoss: 5.0134 (4.3362) TLoss: 0.6928 (0.6920) KLoss: 2.8371 (2.9129) Accu: 33.30
Epoch: [7][300/695] Time: 0.274 (0.275) lr:0.080 Loss: 4.8394 (4.9915) iLoss: 4.1478 (4.2997) TLoss: 0.6916 (0.6918) KLoss: 3.1658 (2.9507) Accu: 34.38
Epoch: [7][350/695] Time: 0.273 (0.274) lr:0.080 Loss: 5.1756 (4.9720) iLoss: 4.4841 (4.2805) TLoss: 0.6915 (0.6915) KLoss: 3.5287 (2.9975) Accu: 35.17
Epoch: [7][400/695] Time: 0.272 (0.274) lr:0.080 Loss: 4.3831 (4.9398) iLoss: 3.6960 (4.2486) TLoss: 0.6871 (0.6912) KLoss: 3.2113 (3.0200) Accu: 36.13
Epoch: [7][450/695] Time: 0.273 (0.274) lr:0.080 Loss: 5.0209 (4.9078) iLoss: 4.3340 (4.2169) TLoss: 0.6869 (0.6909) KLoss: 3.7779 (3.0700) Accu: 37.02
Epoch: [7][500/695] Time: 0.274 (0.274) lr:0.080 Loss: 4.8175 (4.8818) iLoss: 4.1292 (4.1913) TLoss: 0.6883 (0.6906) KLoss: 3.9559 (3.1194) Accu: 37.55
Epoch: [7][550/695] Time: 0.277 (0.274) lr:0.080 Loss: 4.2501 (4.8495) iLoss: 3.5702 (4.1594) TLoss: 0.6799 (0.6901) KLoss: 3.4374 (3.1671) Accu: 38.31
Epoch: [7][600/695] Time: 0.274 (0.274) lr:0.080 Loss: 4.5579 (4.8126) iLoss: 3.8681 (4.1232) TLoss: 0.6898 (0.6895) KLoss: 4.0766 (3.2236) Accu: 39.06
Epoch: [7][650/695] Time: 0.278 (0.274) lr:0.080 Loss: 4.2724 (4.7809) iLoss: 3.5864 (4.0918) TLoss: 0.6860 (0.6890) KLoss: 3.7125 (3.2726) Accu: 39.65
==> Preparing Data Loader...
8
[10436 10428 10497 ... 16615 16608 16626]
[4488 4496 4490 ... 7152 7141 7142]
Epoch: [8][0/695] Time: 0.705 (0.705) lr:0.090 Loss: 4.5153 (4.5153) iLoss: 3.8360 (3.8360) TLoss: 0.6793 (0.6793) KLoss: 4.3656 (4.3656) Accu: 42.71
Epoch: [8][50/695] Time: 0.273 (0.282) lr:0.090 Loss: 4.7377 (4.2681) iLoss: 4.0614 (3.5888) TLoss: 0.6764 (0.6793) KLoss: 4.0565 (4.2724) Accu: 49.27
Epoch: [8][100/695] Time: 0.273 (0.277) lr:0.090 Loss: 4.3135 (4.2672) iLoss: 3.6423 (3.5883) TLoss: 0.6712 (0.6789) KLoss: 3.9947 (4.1619) Accu: 48.81
Epoch: [8][150/695] Time: 0.273 (0.276) lr:0.090 Loss: 3.4615 (4.2268) iLoss: 2.7908 (3.5479) TLoss: 0.6707 (0.6789) KLoss: 4.0730 (4.2252) Accu: 48.83
Epoch: [8][200/695] Time: 0.272 (0.275) lr:0.090 Loss: 4.0212 (4.1892) iLoss: 3.3501 (3.5113) TLoss: 0.6711 (0.6779) KLoss: 4.3166 (4.2836) Accu: 49.45
Epoch: [8][250/695] Time: 0.273 (0.275) lr:0.090 Loss: 4.3676 (4.1472) iLoss: 3.7044 (3.4704) TLoss: 0.6632 (0.6768) KLoss: 4.7292 (4.3411) Accu: 49.94
Epoch: [8][300/695] Time: 0.276 (0.275) lr:0.090 Loss: 4.1010 (4.1313) iLoss: 3.4423 (3.4550) TLoss: 0.6587 (0.6763) KLoss: 4.7060 (4.3611) Accu: 50.03
Epoch: [8][350/695] Time: 0.272 (0.274) lr:0.090 Loss: 4.1517 (4.0975) iLoss: 3.4715 (3.4221) TLoss: 0.6802 (0.6754) KLoss: 4.9141 (4.4090) Accu: 50.49
Epoch: [8][400/695] Time: 0.273 (0.274) lr:0.090 Loss: 3.8435 (4.0659) iLoss: 3.1813 (3.3920) TLoss: 0.6623 (0.6739) KLoss: 5.2494 (4.4566) Accu: 51.01
Epoch: [8][450/695] Time: 0.273 (0.274) lr:0.090 Loss: 3.6860 (4.0329) iLoss: 3.0416 (3.3602) TLoss: 0.6444 (0.6728) KLoss: 4.6557 (4.5093) Accu: 51.46
Epoch: [8][500/695] Time: 0.273 (0.274) lr:0.090 Loss: 3.7771 (3.9927) iLoss: 3.1474 (3.3220) TLoss: 0.6297 (0.6707) KLoss: 4.8703 (4.5578) Accu: 52.09
Epoch: [8][550/695] Time: 0.273 (0.274) lr:0.090 Loss: 3.3321 (3.9598) iLoss: 2.6735 (3.2901) TLoss: 0.6586 (0.6698) KLoss: 5.0266 (4.6114) Accu: 52.57
Epoch: [8][600/695] Time: 0.273 (0.274) lr:0.090 Loss: 3.7681 (3.9360) iLoss: 3.1343 (3.2677) TLoss: 0.6338 (0.6683) KLoss: 5.0208 (4.6551) Accu: 52.88
Epoch: [8][650/695] Time: 0.274 (0.274) lr:0.090 Loss: 3.3970 (3.9080) iLoss: 2.7462 (3.2409) TLoss: 0.6507 (0.6671) KLoss: 5.3225 (4.7055) Accu: 53.19
Test Epoch: 8
Extracting Gallery Feature...
Extracting Time: 0.808
Extracting Query Feature...
Extracting Time: 4.322
Evaluation Time: 4.040
POOL: Rank-1: 3.29% | Rank-5: 11.78% | Rank-10: 22.09%| Rank-20: 40.84%| mAP: 5.78%| mINP: 3.18%
FC: Rank-1: 2.84% | Rank-5: 11.73% | Rank-10: 19.88%| Rank-20: 35.02%| mAP: 5.09%| mINP: 2.43%
Best Epoch [4]
==> Preparing Data Loader...
9
[21687 21680 21688 ... 20999 20988 20992]
[11329 11337 11337 ... 10667 10661 10658]
Epoch: [9][0/695] Time: 0.697 (0.697) lr:0.100 Loss: 3.1833 (3.1833) iLoss: 2.5300 (2.5300) TLoss: 0.6533 (0.6533) KLoss: 5.8047 (5.8047) Accu: 59.90
Epoch: [9][50/695] Time: 0.273 (0.281) lr:0.100 Loss: 4.1611 (3.5205) iLoss: 3.5199 (2.8762) TLoss: 0.6411 (0.6443) KLoss: 5.5921 (5.4184) Accu: 58.81
Epoch: [9][100/695] Time: 0.272 (0.277) lr:0.100 Loss: 3.3032 (3.4773) iLoss: 2.6719 (2.8344) TLoss: 0.6313 (0.6429) KLoss: 5.8251 (5.4112) Accu: 59.59
Epoch: [9][150/695] Time: 0.273 (0.276) lr:0.100 Loss: 3.6473 (3.4583) iLoss: 3.0196 (2.8168) TLoss: 0.6276 (0.6415) KLoss: 5.5682 (5.4526) Accu: 59.38
Epoch: [9][200/695] Time: 0.273 (0.275) lr:0.100 Loss: 3.9288 (3.4246) iLoss: 3.2923 (2.7851) TLoss: 0.6365 (0.6395) KLoss: 6.0866 (5.5165) Accu: 59.68
Epoch: [9][250/695] Time: 0.271 (0.275) lr:0.100 Loss: 2.9349 (3.4116) iLoss: 2.2829 (2.7739) TLoss: 0.6520 (0.6377) KLoss: 5.3323 (5.5485) Accu: 59.81
Epoch: [9][300/695] Time: 0.273 (0.275) lr:0.100 Loss: 3.1694 (3.3783) iLoss: 2.5291 (2.7414) TLoss: 0.6403 (0.6369) KLoss: 6.2431 (5.6114) Accu: 60.21
Epoch: [9][350/695] Time: 0.272 (0.274) lr:0.100 Loss: 3.1968 (3.3368) iLoss: 2.5118 (2.7021) TLoss: 0.6850 (0.6347) KLoss: 6.4518 (5.6803) Accu: 60.68
Epoch: [9][400/695] Time: 0.272 (0.274) lr:0.100 Loss: 3.2538 (3.3003) iLoss: 2.5923 (2.6681) TLoss: 0.6615 (0.6322) KLoss: 6.4434 (5.7386) Accu: 61.05
Epoch: [9][450/695] Time: 0.272 (0.274) lr:0.100 Loss: 2.1155 (3.2642) iLoss: 1.4993 (2.6341) TLoss: 0.6162 (0.6301) KLoss: 6.0223 (5.7934) Accu: 61.39
Epoch: [9][500/695] Time: 0.273 (0.274) lr:0.100 Loss: 3.0302 (3.2474) iLoss: 2.4420 (2.6189) TLoss: 0.5882 (0.6285) KLoss: 6.1151 (5.8352) Accu: 61.58
Epoch: [9][550/695] Time: 0.276 (0.274) lr:0.100 Loss: 2.0192 (3.2102) iLoss: 1.4496 (2.5842) TLoss: 0.5697 (0.6260) KLoss: 6.1015 (5.8877) Accu: 62.01
Epoch: [9][600/695] Time: 0.272 (0.274) lr:0.100 Loss: 2.4337 (3.1895) iLoss: 1.8641 (2.5652) TLoss: 0.5696 (0.6243) KLoss: 6.1188 (5.9480) Accu: 62.24
Epoch: [9][650/695] Time: 0.272 (0.274) lr:0.100 Loss: 2.9384 (3.1669) iLoss: 2.3256 (2.5440) TLoss: 0.6127 (0.6229) KLoss: 6.9302 (5.9854) Accu: 62.50
==> Preparing Data Loader...
10
[21001 20988 20987 ... 22193 22196 22183]
[10655 10658 10657 ... 11833 11839 11837]
Epoch: [10][0/695] Time: 0.702 (0.702) lr:0.100 Loss: 2.7923 (2.7923) iLoss: 2.2307 (2.2307) TLoss: 0.5615 (0.5615) KLoss: 6.4560 (6.4560) Accu: 64.58
Epoch: [10][50/695] Time: 0.272 (0.282) lr:0.100 Loss: 3.3721 (2.7971) iLoss: 2.7299 (2.2005) TLoss: 0.6422 (0.5966) KLoss: 6.8163 (6.7191) Accu: 66.97
Epoch: [10][100/695] Time: 0.272 (0.277) lr:0.100 Loss: 2.7034 (2.7808) iLoss: 2.1790 (2.1863) TLoss: 0.5244 (0.5945) KLoss: 6.1185 (6.6961) Accu: 67.37
Epoch: [10][150/695] Time: 0.274 (0.276) lr:0.100 Loss: 2.6545 (2.7674) iLoss: 2.0719 (2.1738) TLoss: 0.5826 (0.5936) KLoss: 6.6824 (6.6773) Accu: 67.16
Epoch: [10][200/695] Time: 0.273 (0.275) lr:0.100 Loss: 2.6527 (2.7266) iLoss: 2.0979 (2.1376) TLoss: 0.5548 (0.5889) KLoss: 6.5861 (6.7291) Accu: 67.72
Epoch: [10][250/695] Time: 0.273 (0.275) lr:0.100 Loss: 2.1370 (2.7083) iLoss: 1.6269 (2.1193) TLoss: 0.5101 (0.5890) KLoss: 6.2336 (6.7819) Accu: 68.05
Epoch: [10][300/695] Time: 0.274 (0.275) lr:0.100 Loss: 2.9291 (2.6769) iLoss: 2.3844 (2.0909) TLoss: 0.5447 (0.5860) KLoss: 6.7695 (6.8256) Accu: 68.38
Epoch: [10][350/695] Time: 0.273 (0.274) lr:0.100 Loss: 2.0951 (2.6431) iLoss: 1.5120 (2.0592) TLoss: 0.5831 (0.5840) KLoss: 7.3683 (6.8784) Accu: 68.75
Epoch: [10][400/695] Time: 0.274 (0.274) lr:0.100 Loss: 2.2369 (2.6184) iLoss: 1.6413 (2.0366) TLoss: 0.5956 (0.5818) KLoss: 7.2489 (6.9032) Accu: 68.97
Epoch: [10][450/695] Time: 0.273 (0.274) lr:0.100 Loss: 2.5862 (2.5974) iLoss: 2.0108 (2.0186) TLoss: 0.5754 (0.5788) KLoss: 7.3194 (6.9532) Accu: 69.16
Epoch: [10][500/695] Time: 0.272 (0.274) lr:0.100 Loss: 2.1384 (2.5806) iLoss: 1.6435 (2.0031) TLoss: 0.4949 (0.5775) KLoss: 6.8980 (6.9729) Accu: 69.40
Epoch: [10][550/695] Time: 0.273 (0.274) lr:0.100 Loss: 1.9845 (2.5613) iLoss: 1.4426 (1.9861) TLoss: 0.5419 (0.5753) KLoss: 7.8287 (7.0213) Accu: 69.57
Epoch: [10][600/695] Time: 0.274 (0.274) lr:0.100 Loss: 2.3351 (2.5333) iLoss: 1.7581 (1.9603) TLoss: 0.5771 (0.5731) KLoss: 7.4241 (7.0556) Accu: 69.93
Epoch: [10][650/695] Time: 0.273 (0.274) lr:0.100 Loss: 2.3799 (2.5136) iLoss: 1.8113 (1.9418) TLoss: 0.5686 (0.5718) KLoss: 7.2985 (7.0955) Accu: 70.18
Test Epoch: 10
Extracting Gallery Feature...
Extracting Time: 0.774
Extracting Query Feature...
Extracting Time: 4.337
Evaluation Time: 4.122
POOL: Rank-1: 5.15% | Rank-5: 18.22% | Rank-10: 31.03%| Rank-20: 51.04%| mAP: 8.47%| mINP: 4.90%
FC: Rank-1: 3.76% | Rank-5: 14.36% | Rank-10: 25.45%| Rank-20: 41.65%| mAP: 6.22%| mINP: 3.08%
Best Epoch [4]
==> Preparing Data Loader...
11

[cq0907]

Hi, can you upload the training log for the first 10 epochs?

mAP curve:

rank1 curve:

mINP curve:

[zesenwu23]

You can try to change the code of line 148 in ICCV21_CAJ/loss.py,
from diff_pow =torch.clamp_max(diff_pow, max=88)
to diff_pow =torch.clamp_max(diff_pow, max=44).

[cq0907]

OK, thank you. I'll try again.

[cq0907]

Let me ask you one more question. What's the difference between the following two? I'm a beginner. Please forgive me.

[zesenwu23]

Let me ask you one more question. What's the difference between the following two? I'm a beginner. Please forgive me.

You may refer to this issue FC or Pool feature?

[cq0907]

Let me ask you one more question. What's the difference between the following two? I'm a beginner. Please forgive me.

You may refer to this issue FC or Pool feature?

OK, thank you!

[xiaoye-hhh]

I change the code

            # diff_pow =torch.clamp_max(diff_pow, max=88)
            diff_pow =torch.clamp_max(diff_pow, max=44)

but the result is still low. What should i do？

[cleddy]

It happened the same to me. The results reach only 20%+ in mAP of my best epoch.I found that the triplet loss drop dramatically in epoch 5, just like yours.

I don't know whether it is right.

[intlabSeJun]

I found reason. enhanced Tripletloss. if u change the Triplet loss 'agw' mode then it good working. but best result will not reproduce.

[intlabSeJun]

like this. use AGW tripletloss. but args.method='adp'

[ling275]

I also have this problem. Improvements have been made in accordance with the recommendations, but the results are still not met. Have you solved this problem?