Python SDK for Product detection on shelf using Computer Vision and custom vision.

Rishi Khemka 60

Hi im unable to find the correct documentation for product detection on shelf, using both computer vision and also using custom vision. I want to know whether a documentation exists outlining all functions and parameters that go into each function using python.

Also, if i could find any sample documents implementing the same, please let me know.

Thank you

santoshkc 4,425 Reputation points Microsoft Vendor

2024-04-24T08:38:29.64+00:00
Hi @Rishi Khemka,

Thank you for reaching out to Microsoft Q&A forum!

To perform product detection on shelf using Computer Vision and Custom Vision in Microsoft Azure Cloud, you can use the Azure Cognitive Services Computer Vision API and Custom Vision Service.

Here are the links to the documentation:

Computer Vision API

Custom Vision Service

To get started with the Computer Vision API, you can use the Python SDK for Computer Vision. Here's a link: Quickstart: Computer Vision client library for Python.

To get started with the Custom Vision Service, you can use the Python SDK for Custom Vision. Here's a link: Quickstart: Custom Vision client library for Python.

I hope this information helps! Thank you.
santoshkc 4,425 Reputation points Microsoft Vendor

2024-04-25T08:00:21.48+00:00

Hi @Rishi Khemka,

Following up to see if the given response was helpful. Thank you.
Rishi Khemka 60 Reputation points

2024-04-25T08:50:20.84+00:00

Hi @santoshkc i'm looking for the exact python sdk for production recognition using Computer Vision and product recognition using Custom Vision both

santoshkc 4,425 Microsoft Vendor

Hi @Rishi Khemka,

Python sdk for product recognition using Computer Vision:

from azure.cognitiveservices.vision.computervision import ComputerVisionClient
from azure.cognitiveservices.vision.computervision.models import VisualFeatureTypes
from msrest.authentication import CognitiveServicesCredentials

# Replace with your subscription key and endpoint
subscription_key = "<your-subscription-key>"
endpoint = "<your-endpoint>"

# Create an instance of the ComputerVisionClient
computervision_client = ComputerVisionClient(endpoint, CognitiveServicesCredentials(subscription_key))

# URL of the image containing the product
image_url = "https://example.com/image.jpg"

# Specify the features to extract
features = [VisualFeatureTypes.tags]

# Perform image analysis
image_analysis = computervision_client.analyze_image(image_url, visual_features=features)

# Extract tags from the image
if image_analysis.tags:
    print("Tags detected in the image:")
    for tag in image_analysis.tags:
        print(tag.name)
else:
    print("No tags detected in the image.")

Output snippet:

User's image

Use the computer vision for product detection to get the better results.

Custom vision is used for object detection and here is python sdk link for custom vision: Create an object detection project with the Custom Vision.

I hope this helps. Thank you.

Rishi Khemka 60 Reputation points

2024-04-26T05:16:16.6666667+00:00

Hi @santoshkc thank you for the detailed response, the idea is on the shelf i want to be able to identify only Magazine X, there will be various magazines kept, but i need to specifically identify magazine X.

Also Magazine X releases a new version every week and the letters of magazine X are not always visible due to individuals standing in front of the magazine, say the magazine name is people magazine, so sometimes opl will be done, sometimes peo will be hidden, with a new face every week, so will this suffice my use case? which service should be exactly using Computer vision or custom vision ?
Also please provide the relevant sdk code and if possible the model customization code as well if you suggest to go ahead with computer vision

Thanks a lot for all the help
santoshkc 4,425 Reputation points Microsoft Vendor

2024-04-26T08:47:13.27+00:00

Hi @Rishi Khemka,

Thank you for the additional information. To identify a specific magazine on a shelf, you can use the Computer Vision API with object detection. You can train a custom object detection model to recognize the specific magazine you are interested in. You can try using above suggested code.

For more info see: Quickstart: Computer Vision client library for Python. Try to adjust the code as per your use case.

I hope it helps. Thank you.