Industry safety application (BA3) is a project focused on hazard detection in industrial conditions by real-time recognizing the presence or the absence of required safety equipment in people.
Object detection is achieved by fine tuning a pretrained object detection algorithm, YOLOv5 specifically, with a synthetic dataset generated specifically for this project.
The process of creating the synthetic dataset benefits from the features of the Unity Game Platfrom and the information about 3D object that can be extracted from it.
This repository is separated into two different directories
1. synthetic-dataset-generator
This is the Unity Project. It contains all necessary configurations to create an annotated synthetic dataset for fine tuning.
This projects contains the necessary code to fine tune the object detection algorithm. It also provides a cli program that inference real-time the camera feed to the trained model and creates an alert if the required equipment is missing.
This small project contains code to transform the generated dataset of the Synthetic Dataset Generator to an acceptable format for the Object Detection algorithm that was used for training
More information about each subproject, step-by-step instructions, installation guidelines etc., are available on the project's wiki page.
All licences of the 3D models used for the creation of the dataset are provided in a LICENCE file