Mr.Hang
Mr.Hang
For smpl mesh, the simplest way is to use the trimesh API: ```python #for body curvature_mesh = trimesh.curvature.discrete_gaussian_curvature_measure(mesh, vertices, 0.12)[:,None]. #for hand curvature_mesh = trimesh.curvature.discrete_mean_curvature_measure(mesh, vertices, 0.06)[:,None] ``` For object...
Sorry for the late reply, we have already started uploading the latest results to Google Drive.
Hi, please use this [line](https://github.com/paulchhuang/bstro/blob/d67647ae4e05d2ee6ec919e42470cd558a89d247/scripts/download_models.sh#L20) to download
We have tested some newly released models, like kling1.6, tongyi2.1, etc. with our refined prompts, we will upgrade them in our VGenEval2.0
> Thank you for the great work done here! Will Tongyi T2V 2.1 be included in the comparisons? Wan2.1 (T2V and I2V) are included now
We will add it when Veo2 is ready for public testing
1. smplx -> smplh, see [here](https://github.com/vchoutas/smplx/tree/main/transfer_model) 2. Regarding the shape, both 10 and 16 are ok;
1. Yes, you can use different colors to label the SMPL-H; 2. Use CloudCompare to label the point cloud;
Sorry for the inconvenience. We have updated the link, and you can download it now.
The Agent_T2V and Agent_I2V are extended prompts.