Custom Vision

Custom Vision is a computer vision product for building custom image models from labeled examples. It helps users upload images, train a model on their own concepts, and evaluate new images through REST API calls.

AI 이미지 인식

웹사이트 방문

Overview

Custom Vision is a computer vision product for building custom image models from a small set of labeled examples. Its homepage positions the tool as a way to make visual intelligence easier by adapting a model to a specific use case.

The core workflow on the page is straightforward: upload images, train the model on the labels you provide, and evaluate images through REST API calls. The product appears aimed at users who need a custom image classifier or tagger without building the full pipeline from scratch.

Core capabilities

Image upload and tagging

Upload your own labeled images as the starting point for a custom model, or use the product to add tags to unlabeled images faster.

Model training from examples

Teach the system the concepts you care about by training on your labeled examples.

Image evaluation

Use the trained model to evaluate new images after training is complete.

REST API access

Tag images through simple REST API calls, making the model usable in software workflows rather than only in a web interface.

Custom model creation

Focus the model on a unique use case instead of a generic computer vision task.

Common use cases

Build a custom image model
Teams with a small labeled dataset can use the product to create a model tailored to a specific image classification or tagging task.
Prepare training data
Users can speed up labeling by starting with unlabeled images and adding tags before training the model.
Add image tagging to software workflows
After training, developers can call the REST API to tag images from another application or service.
Fit a specialized use case
Organizations with a narrow visual problem can adapt the model to their own concepts rather than using a general-purpose computer vision tool.

Pros and Cons

Pros

Lets you build a model around your own labeled images instead of relying on a generic model.
Supports assisted tagging of unlabeled images as part of the preparation workflow.
Uses a simple three-step flow that is easy to understand from the homepage.
Provides REST API access for using the model in downstream applications.

Cons

The collected pages do not describe pricing, so commercial terms are unknown from this evidence.
The source does not document integrations, SDKs, or broader platform support beyond REST API calls.
Feature details beyond the basic upload, train, and evaluate flow are limited in the available pages.

FAQ

How does Custom Vision work?

Custom Vision uses labeled images to train a custom computer vision model. The source page describes a workflow of uploading images, training on the labels, and then using the model through REST API calls.

Can I use unlabeled images?

The source page says you can bring your own labeled images, or use Custom Vision to quickly add tags to unlabeled images before training. That suggests it supports both manual labeling and assisted tagging, but the page does not describe team review or collaboration features.

What does evaluation output look like?

The source page only confirms that evaluation happens through simple REST API calls. It does not describe return formats, confidence scores, dashboards, or export options.

How is Custom Vision priced?

The source does not provide pricing, plan tiers, or trial details. The pricing URL returns a 404 error, so availability and commercial terms are not stated on the collected pages.