face detection
C++ CNN-based face detection library with zero external dependencies and SIMD acceleration, designed for cross-platform embedding.
Solves the same problem with a different architecture or at a different layer. Expect to rewrite the integration.
Choose it if you want a PyTorch RetinaFace implementation with training code rather than an embeddable dependency-free C++ library.
Retinaface get 80.99% in widerface hard val using mobilenet0.25.
Choose it for a maintained pip-installable Python face detector with alignment, accepting a heavier runtime than the seed.
RetinaFace: Deep Face Detection Library for Python
Choose it if you need an anchor-free CenterFace detector in C++ with landmarks, but it lacks recent maintenance.
face detection
Choose it for a modern YOLOv7-based face detector with landmarks in PyTorch, at the cost of a heavy ML runtime.
yolov7 face detection with landmark
Choose it if you want lightweight YOLO face detection with landmarks and training code, but it is unmaintained since 2020.
yoloface大礼包 使用pytroch实现的基于yolov3的轻量级人脸检测（包含关键点）
Choose it if you need a PyTorch RetinaFace codebase for training your own face detector rather than a ready C++ model.
这是一个retinaface-pytorch的源码，可以用于训练自己的模型。
An earlier or more limited way to do the job. Still the right call when you need something small, proven, or CPU-only.
Older C YOLO-family detector for constrained edge devices; a less accurate, unmaintained alternative to the seed.
MobileNetV2-YoloV3-Nano: 0.5BFlops 3MB HUAWEI P40: 6ms/img, YoloFace-500k:0.1Bflops 420KB:fire::fire::fire:
Older TensorFlow Object Detection API based face detector; only relevant if you are tied to a legacy TF pipeline.
A mobilenet SSD based face detector, powered by tensorflow object detection api, trained by WIDERFACE dataset.
Old C++ RetinaFace port with no maintenance in over seven years; use only if you specifically need that implementation.
RetinaFace detector with C++
A stale PyTorch reimplementation of RetinaFace suited for study rather than production use.
Reimplement RetinaFace with Pytorch
Older darknet-based training/inference framework for face detection with landmarks; framework-specific and stale.
加入关键点的darknet训练框架，轻量级的人脸检测，支持ncnn推理
Narrow MNN-specific implementation of LFFD for edge devices; use it only if your stack is already built on MNN.
MNN Implementation of the paper of LFFD: A Light and Fast Face Detector for Edge Devices
Older MTCNN-based detector with head-pose support; less accurate and slower than the seed's single-pass CNN.
Realtime Face Detection and Head pose estimation on Windows、Ubuntu、Mac、Android and iOS
Legacy Android/JVM face detection helper, obsolete and not a cross-platform C++ alternative.
Face detection for your Android app
Old YOLOv3 face detection demo/script, less maintained and less embeddable than the seed library.
Deep learning-based Face detection using the YOLOv3 algorithm (https://github.com/sthanhng/yoloface)
These projects were analysed and named libfacedetection among their alternatives. The relationship is not symmetric — how libfacedetection rates them is a separate judgement, made when libfacedetection is analysed in its own right.
calls libfacedetection “Use when you want a current, self-contained C++ face detector with 1000 FPS rather than a PyTorch-trainable model with NCNN/MNN exports.”
💎1MB lightweight face detection model (1MB轻量级人脸检测模型)