search

boycehbz / DMMR

The code for 3DV 2021 paper "Dynamic Multi-Person Mesh Recovery From Uncalibrated Multi-View Cameras"
MIT License
78 stars 6 forks source link

pose json version #4

Closed skyhe1410 closed 2 years ago

skyhe1410 commented 2 years ago

In the json file of 2d pose, my version is ‘alphapose v0.3', but why is '1.1' in yours. my alphapose code is the latest on github.

boycehbz commented 2 years ago

it doesn't matter

skyhe1410 commented 2 years ago

However, the number of pose data in the same picture will be less than yours. I will report an error when I run it.

skyhe1410 commented 2 years ago

1648096793(1)

boycehbz commented 2 years ago

You can detect the 2D pose using AlphaPose with halpe format. It contains 26 keypoints.

boycehbz commented 2 years ago

For example

python scripts/demo_inference.py --cfg configs/halpe_26/resnet/256x192_res50_lr1e-3_1x.yaml --checkpoint pretrained_models/halpe26_fast_res50_256x192.pth --indir examples/demo/ --save_img

skyhe1410 commented 2 years ago

ok, thank you. @boycehbz

skyhe1410 commented 2 years ago

I tried your method,but i still reported an error.please help me. thank you. @boycehbz

skyhe1410 commented 2 years ago

I found that your JSON data has 78 values, but mine has only 54 1648113302(1) 1648113302

.

boycehbz commented 2 years ago

You may use an improper AlphaPose command. You can try to save the detected pose with the same format as my demo data.

skyhe1410 commented 2 years ago

Thank you for your reply. The alphapose command I used is correct and the result format is correct, but the data is less than yours. @boycehbz

JACKYLUO1991 commented 2 years ago

@skyhe1410 我和你的问题是一样的，v0.3版本给出的少于作者的点位数量，请问你是如何解决的？

boycehbz commented 2 years ago

https://github.com/boycehbz/AlphaPose/blob/master/getData-image-halpe-track.sh 这边有一个我fork的alphapose，这个脚本估计的结果拆分后，跟我用的是一样的

JACKYLUO1991 commented 2 years ago

@boycehbz 我觉得是我用了--format参数导致以OpenPose格式输出的原因