ntedgi/node-efficientnet

Provides eight pre-trained model variants (B0-B7) with configurable input resolutions to trade accuracy for inference speed on ImageNet's 1000-class object detection task. Implements the compound scaling technique from the EfficientNet paper, achieving state-of-the-art accuracy with significantly fewer parameters than traditional CNNs. Runs in Node.js environments via TensorFlow.js, enabling server-side and edge inference without Python dependencies.

261 stars and 62 monthly downloads. No commits in the last 6 months. Available on npm.