Deepdetect

DeepDetect is an open source deep learning server and platform from Jolibrain for training models and serving them through a REST API. The server is written in C++ and exposes both a REST and a C++ API, so applications can send images, text, audio, video or tabular data and get predictions back without embedding a machine learning framework themselves. It suits engineering teams that need to put computer vision and other deep learning models into production on cloud servers, desktops or embedded devices from a single codebase. Supported tasks include image tagging, object detection, segmentation, OCR, audio, video, text classification, tabular CSV data and time series. The platform adds a web interface with Jupyter notebooks and GPU support for training and managing models, more than 25 pretrained models for training with small datasets, ready to use models for tasks from object detection to OCR and sentiment analysis, and neural network templates tuned for GPU, CPU and embedded hardware. The platform installs with Docker in GPU or CPU variants and serves its interface locally once the containers start. At the time of review the site lists server v0.29.0 and platform v0.26.2. DeepDetect is free open source software, with the source for the server, platform and clients published on GitHub. No pricing is published; enterprise enquiries go to Jolibrain by email. Jolibrain positions the project for enterprise use, so organizations that want support or custom work should expect to arrange it directly with the company rather than through a listed plan.