This post was written by Anzu and the IBM Alpha Zone Accelerator team.
Anzu, a recently founded Israeli start-up, is a blended in-game advertising platform operating across the 3D space. The company has developed an innovative ad technology that renders standard ad units into virtual environments — VR, AR, and 3D video games — and enables it to serve fully blended ads inside the game world, rather than on top of it.
With the high volume of ads delivered every second — from various sources, in different languages and ad formats — it’s challenging to review ads manually and classify their content type. However, the ability to optimize ad content is critical, especially in games designed for kids where publishers must follow strict regulations. So, it was important for Anzu to find a scalable solution that would help game developers understand what content their users are exposed to and make it possible to enhance user experience. This is why Anzu chose Watson Visual Recognition, Watson Speech to Text, and IBM Cloud Functions to build their AI-powered ad compliance tool.
IBM approach and new architecture
With an enormous amount of incoming data, Anzu needed a robust solution to process and leverage it. After Anzu joined the Alpha Zone Accelerator Program, IBM helped them to build AI-powered ad compliance tools. We developed the new architecture based on 3 main components:
- Watson Visual Recognition
- Watson Speech to Text
- IBM Cloud Functions
Building the Anzu architecture around Watson Visual Recognition, Watson Speech to Text, and IBM Cloud Functions enabled them to create a serverless ad compliance environment and scale quickly in a cost-efficient manner.
With Watson Visual Recognition, Anzu can analyze all of the ad creatives to determine how explicit they are and tag them using the Object Detection and explicit content detection classifiers. Watson Visual Recognition analyzes video frames and image banners while the Watson Speech to Text service analyzes audio tracks of videos. The data collected on each ad enables publishers to identify problematic ads, take action, and easily block inappropriate sources.
Building a scalable solution
- Watson APIs gave Anzu immediate access to high-quality machine learning capabilities based on advanced classifiers, with the ability to easily create and use proprietary classifiers in the future.
- Implementing Watson Services enabled Anzu to get a robust AI analysis of the ad creatives within minutes.
- The IBM Cloud Functions technology created a serverless environment that would run the analysis and make calls to the Watson APIs.
- The new architecture helped Anzu build a scalable system within a day that can easily withstand varying traffic loads in a cost-efficient manner!