/yolov3-helmet

Helmet and Mask Detection on Face using YoloV3

Primary LanguageCOtherNOASSERTION

Helment and Mask Detection

I have used Yolov3 as an object detection model to train 4 classes -

  • head,
  • helmet and mask,
  • mask,
  • helmet

I have chosen yolo, because it is quite fast on par with mobilenet-SSD, gives a median accuracy with less data and training time.

Dependencies - refer requirement.txt

  • cycler==0.10.0
  • kiwisolver==1.2.0
  • matplotlib==3.2.1
  • numpy==1.18.4
  • opencv-python==4.2.0.34
  • pandas==1.0.3
  • pyparsing==2.4.7
  • python-dateutil==2.8.1
  • pytz==2020.1
  • six==1.15.0

Model Params

  • batch=64
  • width=416
  • height=416
  • channels=3
  • momentum=0.9
  • decay=0.0005
  • learning_rate=0.001
  • classes=4

Inference

  • To Run Inference: python3 inference.py --image images/0.jpg --config yolov3-helmet.cfg --weights ../backup/yolov3-helmet_20000.weights --names data.names Screenshot