search

ShanghaiTech-IMPACT / 3D-Structure-guided-Network-for-Tooth-Alignment-in-2D-Photograph

[BMVC 2023] 3D Structure-guided Network for Tooth Alignment in 2D Photograph
12 stars 4 forks source link

How to solve the merging problem and fine tune #1

Open DevilANANDGupta opened 2 months ago

DevilANANDGupta commented 2 months ago

How can we improve this model because the facial landmarks are not capable with side view of face. so i want to improve the performance of model as well as can you suggest me for following changes by which i can generate more reliable output. and also for maintain the color of gums.

and please provide me training code. I am waiting since Jan2024

douyl commented 1 month ago

Developing a whole new model to generate post-orthodontic facial photos is a complex undertaking. Our model consists of three modules, one for segmenting the tooth contours (which is easy to replicate), one for generating the well-aligned tooth contours and the other one for generating the facial photos.

The 2nd module requires 3D intraoral scannings for training, while these medical data are sensitive and condidential, containing personal health information that we are obligated to protect. If you wanna generate the side view of face, you must first train a module by rendering the 3D intraoral scannings into the side view, thereby obtaining the tooth contours from this perspective.

I regret that we cannot provide our training code or dataset, since this project is intended for commercial use and we have agreements with collaborators to restrict how we can share and use data. We appreciate your understanding of these constraints.