Author damianvandoom outlines difficulties encountered when using Azure Vision Studio for object detection, particularly in training custom datasets and facing deprecated APIs.

Is Azure Vision Studio Dead?

Author: damianvandoom

Background

The author is working on a proof-of-concept (POC) involving object detection in images. After researching available options, Azure Vision Studio appeared to fit the requirements, and its out-of-the-box capabilities provided satisfactory initial results in identifying relevant objects.

Attempting Custom Dataset Training

As the project progressed, the author attempted to train a machine learning model using a custom dataset. During this step, the system displayed a warning indicating that the API is deprecated. Critically, the platform did not provide an alternative option or clear guidance on how to proceed. The deployment region for the resource is East US.

Additionally, when trying to use the ‘create new dataset’ feature, the process never completed and persistently hung on the screen. A screenshot of this unresolved operation is provided in the post.

Core Issue and Community Question

The author’s central question is whether Azure Vision Studio is still supported and actively maintained, or if users should seek a different tool for custom AI image analysis workflows. The post underlines concerns about API deprecation without obvious migration or upgrade paths, and the lack of feedback or completion when initiating new dataset creation.

Reference Materials

Summary

This post highlights the author’s first-hand experience with Azure Vision Studio’s limitations in custom model training due to apparent API deprecation and unresolved actions within the platform’s UI, and questions the ongoing viability of the service.

This post appeared first on “Reddit Azure”. Read the entire article here