Learning to Reconstruct People in Clothing from a Single RGB Camera

input data & model

• people rotating in frnt of a singla camera
• semantic segmentation and 2D keypoint detection
• avg. pool on multiview data

related work

• well, not quite easy to understand for me

predefined shape model: SMPL

• and keypoint connectivity? confused

loss

• vertex L_2 loss
  • not EMD or chanfer: predefined topology
• vertex loss in posed space
  • supervises pose and shape
• silhouette overlap
• vertex SMPL loss: ?
• loss on pose params
• loss on joints?