C++ facial recognition open cv neural networks

We are looking to build a facial recognition shared library. The core library must be written in C++ due to future testing requirements. There will end up being multiple stages for ongoing work. Stage 1 - Proof of concept - OpenCV utilizing pre-trained DNN models that are available - Load image by filename and detect face - Write binary face template to disc with 128D vectors, write template data (i.e. eye location, UUID) - Finalize gallery by concatenating enrolled templates and merging template data into a manifest - Search gallery with probe image - Generate candidate list from search sorted by Euclidian distance [0-1] and match w/ manifest data - Build C++ shared library so tests can be called via command line and through another test application Environment Requirements - Library must run on 64 bit version of Ubuntu 20.04.3 LTS (Focal Fossa) w/ Linux kernal 5.4.0-91-generic - Library must be in a format that is dynamically linkable using C++17 compiler g++ version 9.3.0 - Run on x86_64 platforms - Use CPU processing (GPU will be down the road) - Run in single threaded mode - Capable of being parallelized via the fork() function - Can read image, detect face, write template within 1 second or less - Templates must be 32KB MAX (smaller the better)