ROMP
ROMP copied to clipboard
Arthur151
Monocular, One-stage, Regression of Multiple 3D People and their 3D positions & trajectories in camera & global coordinates. ROMP[ICCV21], BEV[CVPR22], TRACE[CVPR2023]
a) I was not able to found an notice about the detailed differences and benefits of 'simple-romp' against 'romp'. Only the name implies, that `simple-romp` might be easier to use....
I am missing an notice about the purpose of training `romp` by myself. For example: Are there any benefits of usage, if i train it again by myself instead of...
感谢您提供的ROMP方法！我尝试使用提供的方法，效果很棒！但是在某些场景下出现了人体无法识别的情况。如图中，红色箭头所指被试者在进行大幅前屈动作时。 由于我不是计算机相关专业的，所以想请问一下，这是什么原因呢？是否有解决方法？ 
In image_dataset.py, if we activate shuffle_mode and then the image will be cropped. Now if a person is cropped out of the region, the 2d kps will mostly fall out...
In the paper, it says that fine tuned model shows the best performance at 120Epoch, but after running the v1_hrnet_3dpw_ft.yml file, it showed the best performance at 1 epoch. Did...
Hey @Arthur151, I was able to understand the joint positions and order in SMPL24 and I know that the pose (72,) is the rotation values for 24 joints in SMPL....
有什么原因和优化思路呢？人体模型驱动系统驱动出来的动作就比较自然，我发现很多smplx版本的数字人体，出来的fbx，动画动作都会有抖动和不连贯，不连贯是采样问题？抖动应该是没做动力学限制？
Hi, thanks for your generous sharing of this wonderful work, and I have tried to test on my own video. when only look the mesh that projected onto the image,...
https://github.com/Arthur151/ROMP/blob/623687a37cb7d1ba4538baf1e3c6f65808a36e2c/romp/lib/utils/augments.py#L167-L169 Here, I guess you meant to calculate a bounding box for a single person, based on the visible kpts and then expend the box. However, when you do `leftTop,...