Diabetic Retinopathy Detection (Retina-AI)

The problem

Manual screening for diabetic retinopathy by ophthalmologists is resource-intensive and procedurally complex. With rising diabetic populations, delayed clinical diagnosis frequently results in irreversible vision impairment or blindness.

The approach

Designed and trained a Convolutional Neural Network (CNN) pipeline in PyTorch to automate retinal fundus image grading. Packaged the inference pipeline into a high-performance REST API with FastAPI and containerized it with Docker, enabling lightweight cross-platform web integration independent of client hardware limitations.

Highlights

• Official Bachelor Thesis in Information Technology at Universitas Gadjah Mada (UGM, 2024), advised by Dr.Eng. Silmi Fauziati, S.T., M.T.
• Deep Learning Convolutional Neural Network (CNN) architecture delivering automated fundus image grading.
• Containerized, low-latency REST API inference service powered by FastAPI and Docker.
• Officially indexed and archived in Universitas Gadjah Mada's ETD Academic Research Repository.