YOLO is an object detection network. It can detect multiple objects in an image/video and puts bounding boxes around these objects. This project is for breast detection during surgery using YOLOv5 model on iOS. The main goal of the project is to perform inference using .coreml model and displaying the result (bounding boxes, labels and confidence) on the video feed in real-time.
Create a project in Xcode on your MacOS. Create a new Resource group and plave your .coreml model inside it. That is all the setup you need to begin.
Most of the code is written in ViewController script. I created two ViewController since I had the choice to use CALayers to view my prediction or I could use UIView.
In case of ViewController_CALayers, I added two additional scipt i.e. DisplayBoxes.swift and ShowFPS.swift. These scripts are sued to to display the predictiosn and current FPS on the video preview.
In case of ViewController_UIView, I added VisionView.swift to draw the boxes.
I also make use of Apple's new MainActor.run isntead of Dispatchque.main.async method to run tasks on main thread.
Currently, I am getting around 27 fps on the new Apple Ipad 11 pro (M1) which is quite ok I think.