Multi-view Multi-person SMPLify3D¶
Multi-view Multi-person SMPLify3D
Overview
Argument
Example
Overview¶
This tool could generate multi-view multi-person SMPLData from keypoints3d.
Argument¶
estimator_config:
estimator_config
is the path to aMultiPersonSMPLEstimator
config file. For more details, see docs forMultiPersonSMPLEstimator
and the docstring in code.start_frame:
start_frame
is the index of the start frame.end_frame:
end_frame
is the index of the end frame.bbox_thr:
bbox_thr
is the threshold of the 2d bbox, which should be the same as the threshold used to generate the keypoints3d.keypoints3d_path:
keypoints3d_path
is the path to the keypoints3d file.matched_kps2d_idx:
matched_kps2d_idx
is the matched keypoints2d index from different views, where is generated by code.image_and_camera_param:
image_and_camera_param
is a text file contains the image path and the corresponding camera parameters. Line 0 is the image path of the first view, and line 1 is the corresponding camera parameter path. Line 2 is the image path of the second view, and line 3 is the corresponding camera parameter path, and so on.perception2d_path:
perception2d_path
is the path to the 2d perception data.output_dir:
output_dir
is the path to the directory saving all possible output files, including SMPLData and visualization videos.visualize: By default, visualize is False. Add
--visualize
makes it True and the tool will visualize SMPLData with an orbit camera, overlay SMPL meshes on one view.enable_log_file By default, enable_log_file is False and the tool will only print log to console. Add
--enable_log_file
makes it True and a log file named{smc_file_name}_{time_str}.txt
will be written.
Example¶
Run the tool with visualization.
python tools/mview_mperson_smplify3d.py \
--estimator_config 'configs/modules/core/estimation/mperson_smpl_estimator.py' \
--start_frame 300 \
--end_frame 600 \
--keypoints3d_path 'output/mvpose_tracking/shelf/scene0_pred_keypoints3d.npz' \
--matched_kps2d_idx 'output/mvpose_tracking/shelf/scene0_matched_kps2d_idx.npy' \
--image_and_camera_param 'xrmocap_data/Shelf/image_and_camera_param.txt' \
--perception2d_path 'xrmocap_data/Shelf/xrmocap_meta_test/scene_0/perception_2d.npz' \
--output_dir 'output/mvpose_tracking/shelf/smpl' \
--visualize \
--enable_log_file