Azure AI Custom Vision
An Azure artificial intelligence service and end-to-end platform for applying computer vision to specific domains.
284 questions
Inaccurate Captions Computer Vision API
![](data:image/svg+xml, %3Csvg xmlns='http://www.w3.org/2000/svg' height='64' class='font-weight-bold' style='font: 600 30.11764705882353px "SegoeUI", Arial' width='64'%3E%3Ccircle fill='hsl(169.60000000000002, 9%, 26%)' cx='32' cy='32' r='32' /%3E%3Ctext x='50%25' y='55%25' dominant-baseline='middle' text-anchor='middle' fill='%23FFF' %3EMH%3C/text%3E%3C/svg%3E)
Michael Henderson
0
Reputation points
I am using python to call the Computer Vision API and it is giving me an inaccurate caption for the image I provide.
When testing the same image in vision studio I am getting accurate captions. I'll provide screenshots of my results below.
Vision Studio Result
Python App Result
Python App Code
I would appreciate it if anyone could give insight as to where I should start investigating the issue.
Azure AI Custom Vision
Azure AI services
Azure AI services
A group of Azure services, SDKs, and APIs designed to make apps more intelligent, engaging, and discoverable.
3,413 questions
{count} votes
-
![](data:image/svg+xml, %3Csvg xmlns='http://www.w3.org/2000/svg' height='64' class='font-weight-bold' style='font: 600 30.11764705882353px "SegoeUI", Arial' width='64'%3E%3Ccircle fill='hsl(166.4, 22%, 26%)' cx='32' cy='32' r='32' /%3E%3Ctext x='50%25' y='55%25' dominant-baseline='middle' text-anchor='middle' fill='%23FFF' %3ESC%3C/text%3E%3C/svg%3E)
SriLakshmi C 4,870 Reputation points Microsoft External Staff2025-02-14T12:19:57.9966667+00:00 Hello Michael Henderson,
Greetings and Welcome to Microsoft Q&A! Thanks for posting the question.
I understand that you are encountering discrepancies between the results from the Computer Vision API in your Python application and those in Vision Studio, here are the few steps to consider:
Ensure that you are using an appropriate confidence threshold for the captions. For the Image Analysis 4.0 API, a confidence threshold of
0.0is recommended, while for the Image Analysis 3.2 API, a threshold of0.4is suggested. If your application is filtering out captions based on confidence, it may be excluding valid results.Confirm that you are using the correct version of the API. The Image Analysis 4.0 API has different features and capabilities compared to the 3.2 version. If you are using an older version, consider switching to the latest one for improved accuracy.
If your application is not set to use gender-neutral captions, it may produce inaccurate results. You can enable this feature by setting the
gender-neutral-captionparameter totruein your API request.Ensure that the images you are testing have good quality and clarity. Low-quality images may lead to less accurate captions.
Check if your Azure AI Vision resource is located in a supported region for the features you are using. Some features may not be available in all regions, which could affect the results.
Review the structure of your API request in your Python application. Ensure that you are correctly including the necessary parameters for captioning.
Kindly refer this Image captions (version 4.0).
I Hope this helps. Do let me know if you have any further queries.
Thank you!
-
SriLakshmi C 4,870 Reputation points Microsoft External Staff
2025-02-17T09:40:08.2+00:00 Did you get any chance to check the above response. Thank you!
-
SriLakshmi C 4,870 Reputation points Microsoft External Staff
2025-02-18T12:03:18.0333333+00:00 We haven’t heard from you on the last response and was just checking back to see if you have a resolution yet. In case if you have any resolution please do share that same with the community as it can be helpful to others. Otherwise, will respond with more details and we will try to help.
Thank you!
Sign in to comment
Sign in to answer