Azure AI Custom Vision - Image Classification

What is Azure AI Custom Vision?

Azure AI Custom Vision is a service for training custom computer vision models without deep ML knowledge. You upload images, label them with categories or bounding boxes, and train a model that recognizes your specific objects. The service uses transfer learning to achieve good results with few images.

Custom Vision offers two modes: Image classification assigns an image to one or more categories. Object detection localizes objects in the image with bounding boxes and confidence scores. Trained models can be hosted in the cloud or exported as compact models for edge devices.

Core Features

• Image classification for single-label and multi-label
• Object detection with bounding boxes
• Transfer learning for fast training with few images
• Export for edge deployment (ONNX, TensorFlow, CoreML)
• Iterative training with performance metrics

Typical Use Cases

Manufacturing companies use Custom Vision for automated quality control. Cameras on the production line detect defects, incorrect assembly, or missing components in real-time.

Retailers use object detection for shelf analysis. Models recognize products, out-of-stock situations, and planogram deviations from camera images.

Medical technology companies develop screening tools for skin lesions, X-rays, or microscopic samples. Custom Vision accelerates prototype development.

Benefits

• No ML expertise required
• Training in minutes instead of days
• Flexible deployment options (cloud and edge)
• Continuous improvement through iterative training

Integration with innFactory

As a Microsoft Solutions Partner, innFactory supports you with Azure AI Custom Vision: use case analysis, data strategy, model training, and production integration.