here all the basic informatics are shown about this project.

model stacking flowchart showcasing our proposed pipeline

GlaucFusion: Segmentation-Aware Glaucoma Detection

GlaucFusion is a dual-branch Vision Transformer framework designed for robust and generalizable glaucoma detection. By combining segmentation-aware and global image-based representations, it achieves high accuracy across multiple benchmark datasets.

To Download the research paper draft -----> https://github.com/Pg1910/glaucoma-detection-ai/blob/main/glaucoma_detection_paper.pdf

Features

• Dual-Branch Architecture:
  • Encoder-Only Mask Transformer (EOMT): Leverages optic disc/cup segmentation masks alongside fundus images.
  • Domain-Adaptive DINOv2-ViT-S/14: Dataset-specific classifiers for robust cross-domain performance.
• Quality Control: Laplacian variance filter automatically discards low-quality inputs.
• Domain Mapping: Dataset classification module routes inputs to the correct source domain.
• Dynamic Fusion: Confidence-weighted fusion module combines predictions from both branches based on their proximity to the decision boundary.

---

By fusing complementary features, GlaucFusion offers improved generalizability, interpretability, and reliability for automated glaucoma detection.