OculusAI Vision Diagnostics

A personal exploration into computational vision science and deep learning. Features retinal fundus pathology screening, automated Ishihara plate digit recognition, real-time vision deficiency simulation, and a digital visual acuity screener.

Note: This is a personal research & demonstration project, not intended for formal clinical diagnosis.

Classification Classes

DR, Glaucoma, Cataract, Normal

99.5%

Ishihara Recognition

Multi-font plate validation

< 120ms

Inference Latency

Optimized CPU pipeline

256 x 256

Input Dimensions

Standard fundus resolution

Project Modules

Diagnostic Tools & Simulators

Designed as modular vision screening experiments running neural network inference on CPU runtimes.

Diagnostic Imaging

MobileNetV2

Retinal Pathology Screening

Transfer-learned MobileNetV2 architecture trained on 256x256 fundus imaging for automated classification of Diabetic Retinopathy, Glaucoma, Cataract, and Normal retina.

Circular fundus validationConfidence distributionClinical PDF export

Launch Screening

Color Discrimination

99.5% Accuracy

Ishihara Chromatic Test

Custom 8-layer convolutional neural network trained on 1,400 multi-font Ishihara plates to identify embedded digits and assess Deutan versus Protan color weakness.

4 color vector themesLikelihood ratio analysisStandardized grading

Start Examination

Optometric Modeling

Interactive

Vision Deficiency Simulator

Real-time color transformation pipeline reproducing perception under Protanopia, Deuteranopia, Tritanopia, and Cataract turbidity with split comparison.

Side-by-side comparisonCustom image uploadLens haze emulation

Open Simulator

Functional Assessment

Calibrated

Visual Acuity Screener

Calibrated Tumbling E optotype screener with physical dimension normalization, multi-tier progression, and Snellen equivalent determination.

Screen mm calibrationSnellen 20/200 to 20/15Directional response matrix

Begin Acuity Test

Project Architecture

Built with TensorFlow, Keras, Flask, Next.js, and React. Created by Aditya to explore automated medical image classification and chromatic vision perception.