Object Detection
Uses computer vision for object or defect detection and includes edge deployment capabilities.
🔗Object Detection Repository🔗
Accelerator Description
The detection solution accelerator provides a pre-packaged solution to train, deploy and monitor custom object detection models using the TensorFlow object detection framework within Azure ML.
The aim is to bring State-of-the-art (SOTA) object detection models quickly into production scenarios particularly around the use of defect detection as seen in many quality control scenarios.
Image Recognition aims to recognize and identify objects in images as well as understanding the content and context. TensorFlow object recognition algorithms classify and identify arbitrary objects within larger images. This is usually used in engineering applications such as social networks for photo tagging. By analyzing thousands of photos of trees for example, the technology can learn to identify a tree in an image it has never seen before.
Architecture
Details of the accelerator:
- Leverages the ML Ops accelerator to provide a configurable and re-usable solution accelerator for computer vision detection use-cases.
- Can deploy the computer vision model as a consumable service endpoint in the cloud (Azure).
- Train models using TensorFlow Object Detection API​​​​​​​ leveraging transfer learning with model zoo pre-trained models.
- Uses Azure ML, Azure DevOps and TensorFlow.
Prerequisites
- Access to an Azure subscription
- Access to an Azure DevOps subscription
- Service Principal Account
Accelerator Components
Branching Strategy
