The PCKh performance metric is the percentage of joints with predicted locationsthat are no further than half of the head segment length from the ground truth.
"PCKh total" excludes the pelvis and thorax joints from the calculation,presumably because they are very easy to predict given that the approximateperson center is provided.
About data preprocessing of LSP dataset #24
Thanks! @bearpaw
How do you compute the head size for the final PCKh? The above code doesn't compute it. Do you use the L2 distance between the neck and head joints?
According to your evaluation code https://github.com/bearpaw/pytorch-pose/blob/master/evaluation/eval_PCKh.py, SC_BIAS = 0.6 is used to scale the pre-computed head size for the MPII data. Do the LSP data also require this?
I find this commented line of code# pts[:, 0:2] -= 1 # Convert pts to zero based
in both the LSP and MPII dataloader files. I guess that the annotations provided in the original .mat file are one based. The python language is zero based. What do you think of this problem?
Thanks!
https://github.com/anibali/eval-mpii-pose#pckh
valid.txt https://github.com/umich-vl/给出的验证测试部分
acc epoch loss LR
0.4808 1 0.013783 0.00025
0.6026 2 0.011612 0.00025
0.5474 3 0.011867 0.00025
0.6875 4 0.009999 0.00025
0.7051 5 0.009881 0.00025
0.7322 6 0.009229 0.00025
0.7424 7 0.009207 0.00025
0.7564 8 0.008949 0.00025
0.7666 9 0.008952 0.00025
0.7700 10 0.008808 0.00025
0.7799 11 0.008589 0.00025
0.7860 12 0.008547 0.00025
0.7944 13 0.008294 0.00025
0.7992 14 0.008168 0.00025
0.7981 15 0.008241 0.00025
0.8012 16 0.008086 0.00025
0.8059 17 0.007908 0.00025
0.8142 18 0.007787 0.00025
0.8133 19 0.007735 0.00025
0.8170 20 0.007698 0.00025
0.8218 21 0.007545 0.00025
0.8183 22 0.007631 0.00025
0.8270 23 0.007396 0.00025
0.8249 24 0.007549 0.00025
0.8309 25 0.007332 0.00025
0.8337 26 0.007307 0.00025
0.8328 27 0.007264 0.00025
0.8251 28 0.007501 0.00025
0.8331 29 0.007385 0.00025
0.8367 30 0.007359 0.00025
0.8393 31 0.007148 0.00025
0.8423 32 0.007053 0.00025
0.8405 33 0.007194 0.00025
0.8389 34 0.007185 0.00025
0.8442 35 0.007085 0.00025
0.8464 36 0.007018 0.00025
0.8454 37 0.006985 0.00025
0.8420 38 0.007052 0.00025
0.8483 39 0.006911 0.00025
0.8463 40 0.006965 0.00025
0.8476 41 0.007034 0.00025
0.8494 42 0.006887 0.00025
0.8499 43 0.006872 0.00025
0.8567 44 0.006796 0.00025
0.8553 45 0.006836 0.00025
0.8547 46 0.006759 0.00025
0.8550 47 0.006803 0.00025
0.8524 48 0.006909 0.00025
0.8500 49 0.006881 0.00025
0.8582 50 0.006754 0.00025
0.8556 51 0.006756 0.00025
0.8570 52 0.006718 0.00025
0.8538 53 0.006878 0.00025
0.8568 54 0.006691 0.00025
0.8546 55 0.006709 0.00025
0.8587 56 0.006807 0.00025
0.8597 57 0.006677 0.00025
0.8608 58 0.006686 0.00025
0.8617 59 0.006605 0.00025
0.8598 60 0.006811 0.00025
0.8580 61 0.006734 0.00025
0.8577 62 0.006693 0.00025
0.8619 63 0.006663 0.00025
0.8631 64 0.006715 0.00025
0.8624 65 0.006582 0.00025
0.8623 66 0.006705 0.00025
0.8611 67 0.006707 0.00025
0.8660 68 0.006548 0.00025
0.8597 69 0.006614 0.00025
0.8606 70 0.006597 0.00025
0.8618 71 0.006554 0.00025
0.8612 72 0.006649 0.00025
0.8606 73 0.006679 0.00025
0.8643 74 0.006672 0.00025
0.8671 75 0.006525 0.00025
0.8627 76 0.006650 0.00025
0.8634 77 0.006549 0.00025
0.8660 78 0.006585 0.00025
0.8651 79 0.006523 0.00025
0.8643 80 0.006628 0.00025
0.8650 81 0.006517 0.00025
0.8646 82 0.006577 0.00025
0.8664 83 0.006573 0.00025
0.8653 84 0.006552 0.00025
0.8655 85 0.006692 0.00025
0.8641 86 0.006558 0.00025
0.8646 87 0.006680 0.00025
0.8664 86 0.006573 0.00025
0.8670 87 0.006528 0.00025
0.8665 88 0.006536 0.00025
0.8674 89 0.006432 0.00025
0.8681 90 0.006430 0.00025
0.8661 91 0.006561 0.00025
0.8676 92 0.006499 0.00025
0.8647 93 0.006504 0.00025
0.8691 94 0.006493 0.00025
0.8720 95 0.006466 0.00025
0.8681 96 0.006476 0.00025
0.8634 97 0.006573 0.00025
0.8636 98 0.006505 0.00025
0.8689 99 0.006420 0.00025
0.8673 100 0.006596 0.00025