MobileFace

Why we included this project

MobileFace is a compact face-recognition toolkit aimed at phones and embedded cameras, and it covers more of the pipeline than most repos of its size. Alongside identification embeddings you get detection, landmark localization, pose, alignment, attribute estimation, and tracking, all with CPU inference times in the low milliseconds, so it works without a GPU. The repo ships runnable MXNet/GluonCV and DLib examples for each stage plus an all-in-one script, which makes it easy to see how the pieces fit before you wire up your own stack. LFW benchmark scripts and t-SNE/confusion-matrix visualization tools are included for sanity-checking whether the embeddings actually separate identities. One caveat: this is a research-oriented reference, not a maintained production SDK, so treat it as a starting point for model selection and pipeline design.