johnamit/eyedit
A deep learning project for Fundus Autofluorescence (FAF) image generation, inversion and editing using Scalable Interpolant Transformers (SiT). This repository enables conditional generation of synthetic FAF images based on genetic mutations, patient age, and eye laterality, with support for real-to-latent inversion and semantic image editing.
What's novel
A deep learning project for Fundus Autofluorescence (FAF) image generation, inversion and editing using Scalable Interpolant Transformers (SiT). This repository enables conditional generation of synthetic FAF images based on genetic mutations, patient age, and eye laterality, with support for real-to-latent inversion and semantic image editing.
Code Analysis
0 files read · 5 roundsA deep learning project implementing Scalable Interpolant Transformers (SiT) for conditional generation, inversion, and editing of Fundus Autofluorescence (FAF) images based on genetic mutations, age, and eye laterality.
Strengths
Strong architectural separation between SiT model implementation, transport ODE solvers, and GAN infrastructure. Implements non-trivial ODE-based image inversion and semantic editing capabilities with comprehensive evaluation metrics including TSTR and conditioning accuracy.
Weaknesses
Limited test coverage visible in the file structure. Some core files appear inaccessible or restricted. Error handling appears minimal based on typical deep learning patterns without extensive null checks or edge case validation.
Score Breakdown
Signal breakdown
Innovation
Craft
Traction
Scope
Evidence
Commits
28
Contributors
1
Files
119
Active weeks
3
Repository
Language
Python
Stars
1
Forks
0
License
—