Updated Version of PowerAI Vision by IBM
IBM launched the latest version of its PowerAI Vision product on 15.03.2019. As most of you know, PowerAI Vision is an advanced environment to label data sets for AI models.
Mainly based on IBM internal researches and built upon a number of open-source frameworks, the product is a full-blown Kubernetes implementation. Every new model trained and deployed spawns automatically a number of pre-configured docker containers in order to provide the best possible environment.
One of its inconveniences, especially in multi-tenant implementations for service providers, has been the necessity to occupy a whole GPU for each deployed model in the product.
IBM is promising to have changed this with its latest version – 1.1.3. This is just one of the range of new features in this release:
- Train with a Detectron model
- Transfer learning
- Use non-rectangular shapes when labeling
- Downloadable heat map
We are excited! If you are too, contact L3C and come test the product in our cloud environment!