Face Mask Detection with Machine Learning

Github

: SSD, self model
: MobileNetv2
: Faster RCNN
, : Haar Cascade (OpenCV) + CNN
, : NVIDIA DetectNet_v2 (based on ResNet-18), on Jetson Devices
: SSD (based on Mobilenet and RFB)
: Haar Cascade (OpenCV) + MobileNetv2
: Yolov3 for object detection, Dual Shot Face Detector (DSFD) (better than Haar Cascade) for face detection, ResNet50 for face classification
: RetinaFace (RetinaNetMobileNetV1) for face detection, MobileNetV1 for face classification
: Linzaer for face detection, Paddle Lite for face classification, on Raspberry Pi
: Yolo v2, v3, v4
: ResNet18 on Jetson Nano

Identifying Facemask-Wearing Condition UsingImage Super-Resolution with Classification Networkto Prevent COVID-19, Bosheng Qin : SRCNet

Face Mask with Face Presentation Attack Detection (in this case: mask with part of face), with lighting and distance effect analysis on detection, working on handheld devices, video based

Deep Learning for Face Detection in Real Time
Face Detection : SSD ResNet10 dan MTCNN
Mask Classification : CNN with MobileNetV2 dan VGG16Net
PC and Android Deployment
Variation :
- distance
- lighting
- mask variation (+face attack)
Metric for Performance Analysis :
- Accuracy, Precision, Recall, F1 for image analysis
- mAP@0.5 (Mean Average Precision) for image analysis
- FPS for video analysis
Reference:

Object Detection is Object Localization and Object Classification
Model for Object Detection: Fast R-CNN, Faster R-CNN, Histogram of Oriented Gradients (HOG), Region-based Convolutional Neural Networks (R-CNN), Region-based Fully Convolutional Network (R-FCN), Single Shot Detector (SSD), Spatial Pyramid Pooling (SPP-net), YOLO (You Only Look Once)