🤖Working with AI

A quick guide on implementing and working with AI models

Overview

The following guide will be a brief intro into how to train, convert, upload, and implement AI models into your workflows.

Supported AI Models

Cloneable is built to be able to support a large variety of AI models. Over time we will continue to build out the implementation of various types of frameworks and models.

YoloV8 - Object Detection

Currently we support running pre-trained (by you) yolov8 models. You have the ability to upload the models to the Cloneable platform and use them within the YOLOv8 AI Object Detection component.

Robowflow - Object Detection

We've integrated support to automatically import any models that you train with https://roboflow.com. You will need to input your API key into Cloneable under company settings.

The following page will walk you through utilizing Roboflow models in your workflows

Roboflow Object Detection

How to use YoloV8 models

The following steps will walk you through, at a high level how to take a trained YoloV8 model and get it implemented into a Cloneable workflow

Step 1 - Train a yolov8 object detection model

Follow the guides from the official Ultralytics yolov8 documentation to train an object detection model.

We suggest training a nano, small, or medium model for optimal performance

Step 2 - Convert for iOS

After we have a trained model. You should take the best.pt output and convert it to be able to run on iOS. Soon, we will have an automated conversion process within the Cloneable platform

Below is an example python script to properly convert and optimze your object detection model for Cloneable

from ultralytics import YOLO

# set to the path of your .pt file
# using yolov8n.pt will download the pre-trained yolov8n coco model
model = YOLO('yolov8n.pt') 

# format MUST be set to mlmodel
# nms MUST be True
# half will decrease file size and make more performant
model.export(format="mlmodel", nms=True, half=True)

Once the model is converted, you will find it as yolov8n.mlmodel in the current working directory

Step 3 - Upload model to Cloneable

Navigate to Settings -> Company -> File management

Select upload new file and then select ios_ai_model and select your ml file and then click upload

Your model will now be available in the YoloV8 object detection component

Step 4 - Build the workflow

Now incorporate the uploaded model into your workflow. Here we will discuss basic implementation to show the bounding boxes on a video feed. However, with additional components and logic you can process the bounding boxes to achieve complex tasks such as inspection or counting.

You will utilize the YOLOv8 AI Object Detection component to incorporate your model into your workflow. Once you've dragged the component into the builder, click it to access the list of models which you can choose from to run.

Below is a simple workflow which would run the model and show the bounding boxes over a live camera view.

Example workflow

Below is a guide which walks through building and deploying this workflow

🏗️Building your first workflow