una-dinosauria/3d-pose-baseline

3d output about Leg bending

Closed this issue · 6 comments

  1. my operating system win10
  2. my tensorflow version 1.14
  3. my python version 3.7
    I use openpose to generate two kins of Josns (-BODY_25 and COCO_18) with the same video
    I use the latest fork of https://github.com/ArashHosseini/3d-pose-baseline
    It can output the correct prediction with the COCO_18 json
    test_2

but the _BODY_25 ,the legs are bended
like this
test_2
I want to used the BODY_25 , Faster and more accurate,
i look through the openpose_3dpose_sandbox_realtime.py, it can support the body_25,but the output is not correct

I use this command to predict real time

python src/openpose_3dpose_sandbox_realtime.py --camera_frame --residual --batch_norm --dropout 0.5 --max_norm --evaluateActionWise --use_sh --epochs 200 --load 4874200 --pose_estimation_json output\

but I use this command to predict
python src\openpose_3dpose_sandbox.py --camera_frame --residual --batch_norm --dropout 0.5 --max_norm --evaluateActionWise --use_sh --epochs 200 --load 4874200 --pose_estimation_json output\ --write_gif --gif_fps 24
The predictions are all right about the two kind of jsons

I want to use the openpose_3dpose_sandbox_realtime.py

Did you find any solution @AndroidUnityGit ? Even I have the same issue.