Evaluate Construction Site Safety on iOS using Machine Learning

Building an iOS application for safety on site with Swift, Turi Create, and Core ML

We'll create an iOS application that predicts whether someone is wearing a helmet or not using a simple classifier. But I also added an object detection script if you want to go further.

I'll mainly go through object detection part.

Final result - Classifier

Final result - Object Detection (w/out bounding boxes)

Data preparation

MacOS, Linux & Windows:

./data_preparation > data_preparation.py

Download a large dataset of images with:

• helmet and no jacket
• helmet and jacket
• no helmet and no jacket
• no helmet and jacket

Try to balance out the dataset to improve your results.

I've used an open source repo on GitHub where you can annotate and generate a .csv file with the 7 columns formatted in a way that can easily be used with turicreate.

Simple Image Annotator by Sebastian G. Perez

You should at this point have a .csv file with the necessary information to train an object detection model.

The simple Image Annotator will output a file at the end of the annotation process called out.csv.

Train the model

MacOS, Linux & Windows:

./training_script > main.py

About me

Omar MHAIMDAT

• Linkedin
• Email: [email protected]
• Look at the rest of my repos

Distributed under the MIT license. See LICENSE for more information.