Method | MPJPE (mm) |
---|---|
PoseFormer | 44.3 |
MixSTE | 42.4 |
P-STMO-S | 44.1 |
Diff3DHPE-MixSTE | 42.0 |
Method | MPJPE (mm) |
---|---|
MixSTE | 40.9 |
P-STMO-S | 42.8 |
Diff3DHPE-MixSTE | 40.0 |
Method | MPJPE (mm) |
---|---|
PoseFormer | 31.3 |
MixSTE | 25.9 |
Diff3DHPE-MixSTE | 24.2 |
Method | MPJPE (mm) |
---|---|
MixSTE | 21.6 |
P-STMO-S | 29.3 |
Diff3DHPE-MixSTE | 20.2 |
Method | Frames | PCK (%) | AUC (%) | MPJPE (mm) |
---|---|---|---|---|
PoseFormer | 9 | 88.6 | 56.4 | 77.1 |
MixSTE | 27 | 94.4 | 66.5 | 54.9 |
P-STMO-S | 81 | 97.9 | 75.8 | 32.2 |
Diff3DHPE-MixSTE | 27 | 99.1 | 84.8 | 19.6 |
Please create the envirmennt by the following command:
conda env create -f Diff3DHPE.yml
Please refer to data/README.MD
Please refer to Experiments.sh
Pretrained models OneDirve
python visualization_fig.py --gpu_id 0 -sviz S9 -a "Photo 1" -cam 1 -s 81 -f 81 -b 1 --sampling_timesteps 9 -c checkpoint/h36m/ConditionalDiffusionMixSTES2SGRANDLinLift/cpn/f81 --evaluate ConditionalDiffusionMixSTES2SGRANDLinLift_l2_lr4e-4_useTembed_T_h36m_cpn_81f.bin --config configs/h36m_cpn_s2s_ConditionalDiffusionMixSTES2SGRANDLinLift.json --viz-video data/Videos/S9/Videos/Photo\ 1.55011271.mp4
python visualization_ani.py --gpu_id 0 -sviz S9 -a "Photo 1" -cam 1 -s 81 -f 81 -b 4 --sampling_timesteps 9 -c checkpoint/h36m/ConditionalDiffusionMixSTES2SGRANDLinLift/cpn/f81 --evaluate ConditionalDiffusionMixSTES2SGRANDLinLift_l2_lr4e-4_useTembed_T_h36m_cpn_81f.bin --config configs/h36m_cpn_s2s_ConditionalDiffusionMixSTES2SGRANDLinLift.json --viz-video data/Videos/S9/Videos/Photo\ 1.55011271.mp4 --viz-output viz.mp4
If you find this repo useful, please consider citing our paper:
@InProceedings{Zhou_2023_ICCV,
author = {Zhou, Jieming and Zhang, Tong and Hayder, Zeeshan and Petersson, Lars and Harandi, Mehrtash},
title = {Diff3DHPE: A Diffusion Model for 3D Human Pose Estimation},
booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) Workshops},
month = {October},
year = {2023},
pages = {2092-2102}
}
Our code refers to the following repositories.