search

minar09 / cp-vton-plus

Official implementation for "CP-VTON+: Clothing Shape and Texture Preserving Image-Based Virtual Try-On", CVPRW 2020
https://minar09.github.io/cpvtonplus/
MIT License
353 stars 121 forks source link

What model did you use to generate the human segmentation that includes the neck label? #14

Closed RubReh closed 4 years ago

thaithanhtuan commented 4 years ago

For better results, I would recommend using predicted segmentation by CIHP_PGN pre-trained network, which has a torso-skin label and produces better segmentation prediction. https://github.com/minar09/cp-vton-plus/issues/3#issuecomment-660575377 You can run dataset_neck_skin_correction.py on VITON dataset or use above model as recommended by @minar09

minar09 commented 4 years ago

For better results, I would recommend using predicted segmentation by CIHP_PGN pre-trained network, which has a torso-skin label and produces better segmentation prediction. #3 (comment) You can run dataset_neck_skin_correction.py on VITON dataset or use above model as recommended by @minar09

dataset_neck_skin_correction.py is just a heuristic method, which does not use any pretrained model. It finds the neck area on the existing segmentation and adds the label. For more better training/testing results, CIHP_PGN pre-trained model is better.

minar09 commented 4 years ago

Closing the issue as its resolved. Feel free to reopen in case there are still problems.