yolov8-face VS inference

Hello. Q1: I have a large collection of files from which i have to encode faces. Until now, the fastest way i've found is to use a yolov8 finetune model (https://github.com/derronqi/yolov8-face) for face detection and the face_recognition library for encoding in a 128. I've tried using deepface but it's much slower.

Hijacking this comment because i've been working nonstop on my project thanks to your suggestion. I'm now using this https://github.com/derronqi/yolov8-face for face detection and still the old face_recognition for encodings. I'm clustering with dbscan and extracting frames with ffmpeg with -hwaccel on. I'm planning to try this: https://github.com/timesler/facenet-pytorch as it looks like it would be the fastest thing avaiable to process videos? Keep in mind i need to perform encoding other than just detection because i want to use DBscan (and later also facial recognition, but this might be done separately just by saving the encodings). let me know if you have any other suggestions, and thanks again for your help

Yeah, inference[1] is our open source package for running locally (either directly in Python or via a Docker container). It works with all the models on Universe, models you train yourself (assuming we support the architecture; we have a bunch of notebooks available[2]), or train in our platform, plus several more general foundation models[3] (for things like embeddings, zero-shot detection, question answering, OCR, etc).

We also have a hosted API[4] you can hit for most models we support (except some of the large vision models that are really GPU-heavy) if you prefer.

[1] https://github.com/roboflow/inference

[2] https://github.com/roboflow/notebooks

[3] https://inference.roboflow.com/foundation/about/

[4] https://docs.roboflow.com/deploy/hosted-api