Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Through BrokenEyes: How Eye Disorders Impact Face Detection?

Prottay Kumar Adhikary

TL;DR

A computational framework was developed using the BrokenEyes system to simulate five common eye disorders and reveal critical disruptions in feature maps, particularly for cataract and glaucoma, which align with known neural processing challenges in these conditions.

Abstract

Vision disorders significantly impact millions of lives, altering how visual information is processed and perceived. In this work, a computational framework was developed using the BrokenEyes system to simulate five common eye disorders: Age-related macular degeneration, cataract, glaucoma, refractive errors, and diabetic retinopathy and analyze their effects on neural-like feature representations in deep learning models. Leveraging a combination of human and non-human datasets, models trained under normal and disorder-specific conditions revealed critical disruptions in feature maps, particularly for cataract and glaucoma, which align with known neural processing challenges in these conditions. Evaluation metrics such as activation energy and cosine similarity quantified the severity of these distortions, providing insights into the interplay between degraded visual inputs and learned representations.

Through BrokenEyes: How Eye Disorders Impact Face Detection?

TL;DR

A computational framework was developed using the BrokenEyes system to simulate five common eye disorders and reveal critical disruptions in feature maps, particularly for cataract and glaucoma, which align with known neural processing challenges in these conditions.

Abstract

Vision disorders significantly impact millions of lives, altering how visual information is processed and perceived. In this work, a computational framework was developed using the BrokenEyes system to simulate five common eye disorders: Age-related macular degeneration, cataract, glaucoma, refractive errors, and diabetic retinopathy and analyze their effects on neural-like feature representations in deep learning models. Leveraging a combination of human and non-human datasets, models trained under normal and disorder-specific conditions revealed critical disruptions in feature maps, particularly for cataract and glaucoma, which align with known neural processing challenges in these conditions. Evaluation metrics such as activation energy and cosine similarity quantified the severity of these distortions, providing insights into the interplay between degraded visual inputs and learned representations.
Paper Structure (10 sections, 1 equation, 3 figures, 3 tables, 1 algorithm)

This paper contains 10 sections, 1 equation, 3 figures, 3 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Dataset
  3. BrokenEye: Filter Generation Framework
  4. Statistics
  5. Methodology
  6. Training the Human vs. Non-Human Classifier
  7. Training Disorder-Specific Models
  8. Feature Extraction and Comparison
  9. Result and Discussion
  10. Conclusion

Figures (3)

  • Figure 1: End-to-end experimental pipeline. Raw images from LFW and MS-COCO are processed through the BrokenEyes filter generator to simulate five visual impairments, followed by disorder-specific fine-tuning of ResNet18. Feature maps extracted from layer4 are compared against the normal model using activation energy and cosine similarity to quantify representation drift.
  • Figure 2: Sample data from the dataset. The figure illustrates disorder-specific degradations (AMD, cataract, glaucoma, refractive error, and retinopathy) applied to identical inputs, highlighting characteristic changes such as central scotoma, haze/blur, peripheral vignetting, defocus, and scattered occlusions.
  • Figure 3: Feature map difference visualization. Each heatmap depicts the absolute deviation between the normal model's layer4 activation map and the corresponding disorder-specific model for the same input, highlighting spatial regions where representation drift is concentrated. Warmer colors indicate larger deviations, revealing that cataract and glaucoma produce the most widespread disruptions compared with AMD, refractive error, and retinopathy.