Mission & Technology
Bringing artificial intelligence to the front lines of vision screening and early detection.
👁️ The Vision Behind OJOS AI
Millions of people worldwide suffer from preventable vision loss due to a lack of early screening. Conditions like Glaucoma are famously known as the "silent thief of sight" because they damage the eye gradually without pain. OJOS AI is built to bridge this gap. Our mission is to provide an accessible, immediate, and zero-cost preliminary screening tool that encourages users to seek timely professional medical intervention.
🔬 Deep Learning Core (EfficientNetV2)
Our backend is powered by a state-of-the-art EfficientNetV2-S neural network. The training pipeline consists of:
- CLAHE Preprocessing: Every retinal image undergoes Contrast Limited Adaptive Histogram Equalization to enhance blood vessels and optic disc details.
- Transfer Learning: Fine-tuned on a dataset of 9,800+ high-resolution retinal fundus scans.
- 96.69% Accuracy: Validated rigorously across five classes (Healthy, Cataract, Glaucoma, Diabetic Retinopathy, and Myopia).
By outputting a full probability distribution, users can see the exact breakdown of all possible conditions, rather than a single black-box classification.
💻 Architecture & Tech Stack
We designed the application to be secure, fast, and light. It utilizes the following stack: