Saving Inference Results on VoxelPose

[YoussefFathi]

trafficstars

I have tried running the demo file for the multiview 3D keypoints detection using voxel pose, but could only see the predicted keypoints as visualizations on the images. Is there a way to extract the predicted keypoints themselves in raw JSON format for example?

[wusize]

Hi! You can save the keypoints in .json file in the evaluation process. Add evaluation=dict(res_folder='path/to/save') in the cfg file and run test, the keypoints will be saved under 'path/to/save'.

[fxyQAQ]

Hi, I have added evaluation=dict(res_folder='path/to/save') in the '\configs\body\3d_kpt_mview_rgb_img\voxelpose\panoptic\voxelpose_prn64x64x64_cpn80x80x20_panoptic_cam5.py', but it doesn't work. Is there an update?