DeepStack

Why we included this project

DeepStack runs vision models locally on hardware you control, so you don't have to send images to a cloud API. It serves pre-trained models for face detection and recognition, common-object detection across 80 classes, scene recognition, and 4x image super-resolution through a simple HTTP API. That makes it easy to hook into a home automation hub, a camera system, or an industrial monitoring pipeline without writing model-serving code. The same Docker image works on a Raspberry Pi, an NVIDIA Jetson, a Windows machine, or a GPU server, and everything runs offline, so image data never leaves your network. If the default models don't cover your use case, you can train and deploy custom detectors for objects of your own choosing.