/YoloV3_PPE_prediction

Primary LanguageJupyter Notebook

YOLO v3 PPE(Personal Protection Equipments) Object Detection:

Below PPE objects can be detected from the required video/image

  1. hard hat
  2. vest
  3. mask
  4. boots

PPE object detection after Yolo training with custom dataset:

image

Dataset Preparation:

Custom Dataset was prepared to extract images with PPE objects from Google. 

Code link - https://github.com/rameshveer/YoloV3_PPE_prediction/blob/master/image_scrapper.ipynb

Annotation:

Bounding Boxes(BBox) for the objects in extracted images were manually done using below tool. 

This tool will automatically calculate the x,y co-ordinates of the BBox and put it in yolov3 format via text file.

Training:

After collecting and annotating around 3500 PPE images, they were cleaned and fed to the yolov3 model.

Model was trained using Google Colab(Tesla T4 GPU) for 50 epochs only in 5 hrs. More epochs would fetch far more accurate results.

Youtube link for PPE object detecion with custom Yolo training:

https://www.youtube.com/watch?v=U8lIk9Qhy94

IMAGE ALT TEXT HERE

Yolo V3 architecture can be applied for object detection on various sectors and applications.