Based on paper Understanding 3D Object Articulation in Internet Videos and paper Full-Body Awareness from Partial Observations. Please feel free to point out any mistakes in this repo!
Set up the same environment as Articulation.
Set up additional environment for SMPL and GraphCMR. Make packeges compilable in python 3.7 environment.
To run the model on a video, run
python combined_demo.py --config config/config.yaml --input /z/syqian/articulation_data/step2_filtered_clips/CxTFIEpSgew_34_360.mp4 --output demo_3336_output --save-obj --webvisor
python combined_demo.py --config config/config.yaml --save-obj --webvisTo save the 3d model, add --save-obj and --webvis flags.
Per step demo can be done by running
python generate_mesh_mask.py --config config/config.yaml --input /z/syqian/articulation_data/step2_filtered_clips/CxTFIEpSgew_34_360.mp4 --output /data/siyich/cmr_art/mask_mesh_3336_output --webvisthen
python generate_mesh_mask.py --config config/config.yaml --input /z/syqian/articulation_data/step2_filtered_clips/CxTFIEpSgew_34_360.mp4 --output /data/siyich/cmr_art/mask_mesh_3336_output --webvispython opt_single_img.py --input /data/siyich/cmr_art/mask_mesh_3336_output --frame 60 --output /data/siyich/cmr_art/opt_3336_outputor
python generate_mesh_mask.py --config config/config.yaml --input /z/syqian/articulation_data/step2_filtered_clips/CxTFIEpSgew_34_360.mp4 --output /data/siyich/cmr_art/mask_mesh_3336_output --webvispython opt_pose_img.py --input /data/siyich/cmr_art/mask_mesh_3336_output --frame 60 --output /data/siyich/cmr_art/opt_3336_outputWe reuse the codebase of SparsePlanes and Mesh R-CNN.
