Developing Various Representations.

[ouusan]

1.TexturePose: Supervising Human Mesh Estimation with Texture Consistency(2019) Texture map (texel): A corresponding UVmap un-warps the template surface onto an image, A, which is the texture map code: https://seas.upenn.edu/˜pavlakos/projects/texturepose. (can not open)

2. DenseRaC: Joint 3D Pose and Shape Estimation by Dense Render-and-Compare(2019) IUV map, render and compare scheme, and MOCA synthetic dataset code: No 3.Learning 3D Human Shape and Pose from Dense Body Parts(2020) global/part IUV, Joint-centric RoI Pooling, Part dropout. partial IUV-->rotation features- (graph convolution)->position features-->refined position features(??)-->refined rotation features code: https://github.com/HongwenZhang/DaNet-DensePose2SMPL 4.3D Human Mesh Regression with Dense Correspondence(2020) image-(Correspondence Net)->IUV map-->UV map-(Location Net)->(location map)mesh vertices code: https://github.com/zengwang430521/DecoMR 5.Object-Occluded Human Shape and Pose Estimation from a Single Color Image(2020) human shape estimation is formulated as a UV map inpainting problem. code: No code https://www.yangangwang.com 6.Monocular, One-stage, Regression of Multiple 3D People(ROMP)(2021) Body Center heatmap( integrates the scale information of the body in the 2D image)+Collision Aware representation Mesh Parameter map(consists of Camera map and SMPL map): assuming each location of these maps is the center of a human body, then estimate its corresponding 3D body mesh parameters. Parameter Sampling. code:https://github.com/Arthur151/ROMP 7.Learning to Estimate Robust 3D Human Mesh from In-the-Wild Crowded Scenes(2022) 2D heatmap( crowded scene-robust image feature), 3D heatmap(joint-based regressor), and joint GCN(??) instead of CNN. code:https://github.com/hongsukchoi/3DCrowdNet_RELEASE 8.MotionBERT: A Unified Perspective on Learning Human Motion Representations https://github.com/jr-tagawalab-nttcom/paper-survey/issues/221 a pretraining+task-specific finetune framework. code:https://github.com/Walter0807/MotionBERT

[ouusan]

2.IUV images: estimations on realistic images through DensePose-RCNN https://github.com/facebookresearch/DensePose/blob/main/notebooks/DensePose-RCNN-Texture-Transfer.ipynb

3. (RoI)Spatial transformer networks (graph convolution)Semi-supervised classification with graph convolutional networks
4. image inpainting related
5. related: Putting People in their Place: Monocular Regression of 3D People in Depth(BEV)/2022 TRACE: 5D Temporal Regression of Avatars with Dynamic Cameras in 3D Environments(TRACE)/2023
6. joint specific graph convolution: A comprehensive study of weight sharing in graph networks for 3D human pose estimation.