Azure Custom Vision:Enhancing Vehicle Object Detection with Tailored Models

Introduction

This article will explore how to create a Custom Vision Service in Azure using Object Detection and build and deploy the Azure Custom Vision Service using Object Detection models to detect the different types of vehicles like buses, taxis, and trucks. Using Vehicle Object Detection models will enhance transportation systems and efficiently manage vehicular activities.

What is Object Detection?

Object detection returns the bounding box coordinates (in pixels) for each object found in the image. For example, if an image contains a bus, taxi, or truck, the Detect operation will list those objects with their coordinates in the image.

Azure Custom Vision AI Steps

For creating an Object Detection model using Azure Custom Vision, follow the steps as below.

We have the following steps to create a Custom Vision Service in the Azure Portal.

Go to the Azure portal and sign in with your Azure account.

Search for "Custom Vision" in the search bar and select "Custom Vision" from the search results.

Click on "Create" on the Custom Vision page.

In the Basics tab, provide the following information, and choose Both(Prediction and Training).

Then select your subscription.

Create a new Resource Group as testRG.

Specify the region as East US and enter a name for your Custom Vision service as retailcustomvisionobject.

Select the pricing tier for training resource as Standard.

Similarly, select the pricing tier for prediction resource as Standard.

Click the Review + Create button.

You will get a validation message as Validation Passed.

Click Create button.

Deployment started initializing in a minute or two this became successful.

Steps to follow on the Custom Vision Portal

Step 1. Login into the Custom Vision portal and then click the Sign in button.

Step 2. Custom Vision portal appeared on the screen, and then click Create a new project.

Step 3. Add the images and name the images with proper tagging into a category.

Step 4. Upload the Taxi images to the Add Images button.

Step 5. The users will get a message as images uploaded successfully.

Step 6. Likewise, upload the remaining images, like the bus and truck, to the Add Image button.

Step 7. Now start tagging the appropriate images within the bounded box.

Step 8. Train the model and choose the training type as Quick Training.

Step 9. The users can see three metrics used in the Azure Custom Vision.

Precision: It measures the accuracy of the positive predictions made by the model.

Recall: It measures the model's ability to correctly identify positive instances from the total number of actual positive instances.

mAP: Mean Average Precision assesses precision and recall trade-offs across different thresholds.

Step 10. Test the model either you can upload the images locally or browse the images from the URL.

The final outcome was that the Truck vehicle object was detected successfully with a bounded box of 56.2% probability.

Applications of Custom Vision

There are a variety of use cases for Custom Vision, like,

License Plate Detection - It helps to identify the vehicle license number within images or video frames.

Parking Occupancy Detection - It involves determining the availability or occupancy status of parking spaces in a parking lot.

Pedestrian detection - It helps to detect and track human being movements in various environments such as streets, sidewalks, and roads.

Foot Traffic and People Counting - It helps to analyze video footage or determine the number of people in a particular area.

Summary

In this article, we learned, created, and deployed an Object Detection model to identify vehicles like buses, taxis, or trucks. By leveraging the Azure Custom Vision services and techniques, Vehicle Object detection systems contribute to improving road safety and reducing the number of accidents that happen in the year, which will help future our citizens of India.