liangxuy
Problem with unreasonable test results
Hi, thank you for your great work. I configured the environment as required and downloaded the pre-training model. My directory is as follows: ReGenNet ├── checkpoints │ ├── action_recognition │ │ ├── chi3d_smplx │ │ │ ├── checkpoint_0060.pth.tar │ │ │ └── opt.yaml │ │ └── ntu_smplx │ │ ├── checkpoint_0100.pth.tar │ │ └── opt.yaml │ ├── chi3d │ │ ├── args.json │ │ ├── model000568404.pt │ │ └── opt000568404.pt │ └── ntu │ ├── args.json │ ├── model000147680.pt │ └── opt000147680.pt ├── dataset │ └── chi3d │ ├── annotations_chi3d │ │ ├── chi3d_label.json │ │ └── README.md │ ├── chi3d_smplx_test.h5 │ └── chi3d_smplx_train.h5 ├── myoutput_00 │ ├── rendered（8个动作，每个动作10个结果） │ │ ├── Grab_0.mp4 8 16 24 32 40 48 56 64 72 │ │ ├── Handshake_1.mp4 ...... 73 │ │ ├── Hit_2.mp4 ...... 74 │ │ ├── HoldingHands_3.mp4 ...... 75 │ │ ├── Hug_4.mp4 ...... 76 │ │ ├── Kick_5.mp4 ......77 │ │ ├── Posing_6.mp4......78 │ │ └── Push_7.mp4......79 │ ├── results_len.txt │ ├── results.npy │ └── results.txt ├── actor-x │ └...... ├── assets │ ├── action_names_chi3d.txt │ ├── action_names_ntu.txt │ ├── framework.png │ └── teaser.png ├── body_models │ ├...... ├── cog.yaml ├── data_loaders │ ├...... ├── diffusion │ ├...... ├── docker │ └── Dockerfile ├── en_core_web_sm-2.2.0 │ └── en_core_web_sm-2.2.0.tar.gz ├── environment.yml ├── eval │ ├...... ├── LICENSE ├── model │ ├...... ├── preprocess │ ├...... ├── README.md ├── render │ ├...... ├── sample │ ├── cgenerate.py │ ├── edit.py │ ├── predict.py │ └── pycache │ ├── cgenerate.cpython-312.pyc │ └── cgenerate.cpython-37.pyc ├── train │ ├...... ├── utils │ ├......
Aftert that, I jumped to the "Motion Synthesis and Visualize" step and used the following command to generate and visualize the results.
Generate:
python -m sample.cgenerate --model_path /home/wyj/code/ReGenNet/checkpoints/chi3d/model000568404.pt --action_file /home/wyj/code/ReGenNet/assets/action_names_chi3d.txt --num_repetitions 10 --dataset chi3d --body_model smplx --num_person 2 --pose_rep rot6d --data_path /home/wyj/code/ReGenNet/dataset/chi3d/chi3d_smplx_test.h5 --output_dir /home/wyj/code/ReGenNet/myoutput
Visualize:
python -m render.crendermotion --data_path /home/wyj/code/ReGenNet/myoutput/results.npy --num_person 2 --setting cmdm --body_model smplx
The results didn't look right (see google drive link). I would like to ask you what could be causing this.
I came across the same problem testing Chi3d dataset, have you solved it ?
