fastDRS Models

Model artifacts for the fastDRS diabetic retinopathy screening toolkit.

This repository contains metadata and release artifacts for trained fastDRS models.

Available artifacts

Each registered model may have:

• PyTorch .pth weights
• LiteRT .tflite model

The model binaries are distributed as GitHub Release assets rather than committed directly to Git.

Model architectures

Currently supported:

• ResNet18
• ResNet50
• DenseNet121
• ConvNeXt Tiny
• Swin Transformer Tiny
• MobileNetV2

Downloading models

Models are normally downloaded automatically through fastdrs:

from fastdrs import Predictor

predictor = Predictor.from_pretrained("mobilenet_v2")

For LiteRT:

from fastdrs.inference import LiteRTPredictor

predictor = LiteRTPredictor.from_pretrained("mobilenet_v2")

Downloaded models are cached locally by fastdrs