Feature spotlights

Streamline processes to label, train, monitor and deploy

An intuitive interface helps your work force with no skills in deep learning to compose models for AI solutions. Jobs like labeling and training models are streamlined with technical details abstracted with a series of clicks. Our approach to bring "AI to All" attracts enterprises by driving efficiency and accelerating productivity towards their missions.

Train models to classify images and detect objects

With a few clicks, deep learning models can be trained to classify images or detect objects of importance. Coding to build models is now replaced by simply dragging and dropping images into categories and drawing boundary boxes to tag objects. Technical details like neural networks and hyper parameters are abstracted and pre-configured to learn from sample corpus.

Introducing auto labeling with deep learning models

On average, data scientists spend 80% of their time on labeling and preprocessing data sets for training. In addition to off-loading this task to subject matter experts, we bring iteratively trained deep learning models to automatically label data sets. The resulting data sets add up to build the exhaustive and accurately labeled data sets required for training. Deep learning on data labeling drastically reduces costs and accelerates time for enterprises to deploy AI solutions.

Video analytics made easy for training and inference

In addition to images, our tools can work with videos to create data sets and infer. With a few clicks, your videos can be imported and frames can be processed to label data sets. The trained models can annotate streams of videos with objects.

Extend AI solutions with custom models.

Data scientists can also import custom models (TensorFlow) to be trained, tuned, monitored, and deployed. PowerAI Vision also supports the customization of preprocessing raw images during labeling of data sets. Data scientists can now off load the job of training and deployment to focus on creating innovative models for their missions.

Deploy models on-premises, in the cloud, and on edge devices

PowerAI Vision provides a flexibility with deploying trained models. A central compute intense resource can be allocated for training, but the resulting model can be deployed in local data centers, to the cloud and even on edge devices with AI chips. A clicker tool for developers compiles accelerated models to be deployed onto FPGA cards.

Customer case studies

  • Train models to classify images

  • Auto label videos to train models for object detection

  • Employ continuous learn to label objects

How customers use it

  • Ensuring the safety of workers

    Ensuring the safety of workers


    According to the International Labor Organization every 15 seconds, 151 workers have a work-related accident and 321,000 fatal occupational accidents. Work accidents remain a huge, cross-industry problem, despite safety regulations and procedures.


    Industries are applying AI technologies to monitor and enforce regulations for safety. Embedded computer vision applications can flag workers while entering hazardous environments or scan a construction area to alert supervisors to act.

  • Drone surveillance for Energy and Utilities

    Drone surveillance for Energy and Utilities


    Power companies rely on manpower to visually inspect their towers over large areas. Manual inspections are notorious for being expensive, risky, and slow especially when the towers are spread over mountainous inaccessible terrain.


    Power companies are transforming inspection jobs by deploying drones with cameras to capture inspection data. AI for such industries can help decrease time, increase frequency and reduce risk to workers.

  • Visual insights for quality

    Visual insights for quality


    Manufacturing operations use visual confirmation to ensure t parts have zero defects. The volume of inspections, product SKUs, and the variety of defects pose challenges to delivering a high-quality product.


    Deep learning models deployed on factory floors ensure little decision latency during production. Systems learn continuously by taking feedback from manual inspectors. AI is beginning to deliver reliable results with low escape rates.