flyyy

Release the code of "Motion Guided 3D Pose Estimation from Videos"

2

When do you have the plan to release the code of "Motion Guided 3D Pose Estimation from Videos"?

Training code

3

Thanks for sharing this great and interesting work! I want to know when will you release the training code?

Test on 3DPW-OCC and 3DOH

1

The PARE model can only test on 3DPW, but the paper test on 3DPW, 3DPW-OCC and 3DOH. So how to test the PARE model on 3DPW-OCC and 3DOH datasets?

Question about PARE and HMR

1

In the config, PARE uses "keypoint_dst='smpl_49'" in [hrnet_w32_conv_pare_coco_cache](https://github.com/open-mmlab/mmhuman3d/blob/main/configs/pare/hrnet_w32_conv_pare_coco_cache.py) or "keypoint_dst='smpl_24'" in [hrnet_w32_conv_pare_mix_cache](https://github.com/open-mmlab/mmhuman3d/blob/main/configs/pare/hrnet_w32_conv_pare_mix_cache.py) while HMR uses "keypoint_dst='smpl_54'" in [resnet50_hmr_pw3d_e50_cache](https://github.com/open-mmlab/mmhuman3d/blob/main/configs/hmr/resnet50_hmr_pw3d_e50_cache.py). I want to know what is the difference between them and why...

The results reported by CLIFF in PA-MPJPE on 3DPW are 45.7 mm (Res-50), and 43.0 mm (HR-W48). But the given pretrained model is 47.38 mm (Res-50). And the CLIFF paper...

I install torch1.13.0+cuda116 and mmcv using: ``` conda install pytorch==1.13.0 torchvision pytorch-cuda=11.6 -c pytorch -c nvidia pip install "mmcv-full>=1.3.17,

Camera loss

1

In the [PARE log](https://openmmlab-share.oss-cn-hangzhou.aliyuncs.com/mmhuman3d/models/pare/with_mosh/20220427_113717.log?versionId=CAEQOhiBgMClqr3PgxgiIGRjZWU0NzFhMmVkMDQzN2I5ZmY5Y2MxMzJiZDM3MGQ0), the camera_loss always equal 0.0000, why ?

[CLIFF] 在Human3.6M的评估结果

2

请问论文中的表1在3DPW、Human3.6M、AGORA的评估结果，是用同一个模型测试的吗？ 若不是。我在 https://github.com/huawei-noah/noah-research/issues/94#issuecomment-1204815175 看到在3DPW评估的模型训练数据为[0.4, 0.3, 0.1, 0.2] for [H36M, in-the-wild datasets, 3DHP, 3DPW] ，那么在Human3.6M、AGORA评估的模型，训练数据用的是哪些，如何划分？

[CLIFF] Result on 3DPW without trainning on 3DPW

6

The paper reports the result with training on 3DPW (PA-MPJPE=43.0). But many papers report the result without training on 3DPW, I want to know the result without training on 3DPW?

How can I get the "NoisyAudio/-5dB/MONEY_00581.npz"?

1

When I run the main.py, I get the FileNotFoundError: No such file or directory 'MONEY/NoisyAudio/-5dB/MONEY_00581.npz'