DeepDetect is an open-source deep learning server exposing training and inference through a REST API, splitting into two pieces: a portable Server, and a Platform providing a web UI with Jupyter notebooks and GPU support.
That separation is well judged. The Platform covers the experimentation phase where notebooks and a UI genuinely help, and the Server is the lightweight component you actually deploy – so what runs in production is not carrying the weight of the development environment. Both install via Docker or build from source, and both are versioned openly with public changelogs.
Being open source and self-hosted is the whole proposition: models and data stay on your own infrastructure, with no per-inference pricing and no third party in the request path. The trade is the usual one – you own the operations, the GPU provisioning and the upgrades. This is a tool for teams with the engineering capacity to run it.
