IBM PowerAI Vision overview

The IBM® PowerAI Vision platform, built on cognitive infrastructure, is a new generation of video/image analysis platforms. The platform offers built-in deep learning models that learn to analyze images and video streams for classification and object detection.

PowerAI Vision includes tools and interfaces for anyone with limited skills in deep learning technologies. You can use PowerAI Vision to easily label images and videos that can be used to train and validate a model. The model can then be validated and deployed in customized solutions that demand image classification and object detection.

The following are the main features of PowerAI Vision:

Streamlined model training: You can use existing models that are already trained as starting point to reduce the time required to train models and improve trained results.
Single-click model deployment: After you create a training model, you can deploy an API with one click. You can then develop applications based on the model that you deployed.
Data set management and labeling: You can manage both raw and labeled data.
Video object detection and labeling assistance: Videos that you import can be scanned for objects and the objects can be automatically labeled.

Architecture overview

The architecture of PowerAI Vision consists of hardware, resource management, deep learning computation, service management, and application service layers. Each layer is built around industry-standard technologies. This image illustrates the architecture layers with Vision recognition service on the top, the service management layer next, the computation layer below that, followed by the resource management layer, then the components: POWER, Accelerator, Data store, and Network on the bottom.

Table 1. Overview of the architecture layers
Architectural Layer	Description
Infrastructure Layer	Consists of hardware systems that support PowerAI Vision, including virtual machines (containers), accelerators (GPUs/FPGAs), storage systems, networks, and so on.
Resource Management Layer	Coordinates and schedules all computing resources.
Deep Learning Calculation Layer	Consists of deep learning algorithms, including data processing modules, model training modules, and prediction modules.
Service Management Layer	Manages user projects in a graphical interface, including image preprocessing, data annotation management, data set management, training task management, model management, and API management.
Application Service Layer	Located on the top of the PowerAI Vision platform, it is responsible for managing all application-related services, including image labeling and preprocessing services, video annotation services, customized image classification services, and customized object detection services.