DeepStack
An open-source computer vision API that runs offline on your hardware, with face recognition, object detection and support for custom models.
DeepStack is a self-hosted computer vision API for developers adding image analysis to camera systems, home automation or other applications. It runs prebuilt and custom models on your own hardware and works fully offline. Image processing stays on the device or server where you host it, with no cloud service required.
Its face APIs detect faces, recognize registered people and compare faces for matches. Object detection covers common objects, while custom models let you train and deploy detectors for objects specific to your application. Scene recognition identifies image scenes, and image superresolution enlarges images by a factor of four.
DeepStack runs through Docker on Linux and macOS, as well as NVIDIA Jetson, Raspberry Pi and ARM64 devices. Windows 10 has a native application with CPU and GPU support; Linux supports CPUs and NVIDIA GPUs. You can also host it on a cloud virtual machine running Docker, where processing takes place on that server rather than your local device.
The surrounding ecosystem includes Home Assistant add-ons for object, face and scene recognition, plus integrations with Blue Iris and Agent DVR. Node-RED support and an automation integration with MQTT and Telegram connect its results to other systems. These connections make it relevant to monitoring and IoT projects that need image analysis within an existing setup.
DeepStack’s source code uses Apache-2.0. Custom and third-party model weights have their own licenses. Its API supports SSL and API keys to protect access to your server.
