Open zhudu opened 8 months ago
Hi, currently we have not cleaned the code for video generation. One quick idea is to set the SMPL pose to be the specific pose you hope to use and then render images with these poses under the camera poses. After you get the images, you can convert these images into a video.
According to the steps in readme, I can use the model and data for inference to generate images with different poses. I wonder if you have provided sample code for video generation? I tested gen_ Videos.py uses the following command, but this is not feasible. Is there a demo of inference for generating videos?
‘python gen_videos.py --outdir=logs/exp/ --seeds=0-31 --grid=4x2 --network=logs/training-RenderPeople-runs-450-subject-rs-512-1d-2d-3d-feature-NeRF-decoder-use-trans-sample-obs-view-gpu-4/SHERF_RenderPeople.pkl’