Since we're already linking against TensorFlow and want to keep the number of dependencies small, we should investigate alternative approaches in addition to the obvious solution to use dlib (which is the popular/standard way, see go-face). Pigo is a pure Go implementation for Face Detection, but it can not do Face Recognition.
- https://cheppers.com/deploying-distributed-face-recognition-application-kubernetes - Deploying a distributed face-recognition application with Kubernetes
- http://dlib.net/ - Machine learning library, good for face recognition
- Kagami/go-face - implements face recognition for Go using dlib
- https://hackernoon.com/face-recognition-with-go-676a555b8a7e - face recognition howto using Kagami/go-face
- esimov/pigo - Pigo is a face detection library implemented in Go
- https://gocv.io/ - GoCV gives access to the OpenCV 4 computer vision library
- davidsandberg/facenet - Face Recognition using Tensorflow (Python, not Go)