Google open sources image recognition AI model, VR audio tools

Direct link to the github project:

https://github.com/tensorflow/models/tree/master/research/de...

And the blog post of the announcement:

https://research.googleblog.com/2018/03/semantic-image-segme...

Google Resonance looks really interesting. It's similar to Steam Audio, but AFAIK Steam Audio's core is not open source and Resonance should be more mobile-friendly due to optimization/compromises in head-related transfer function computation. Unfortunately both Steam Audio and Resonance focus on just 3D spatialization and skip on basic functionality like pitch shifting.

It is a amazing how much cool stuff Google gives away.

It did a really poor job on the pictured cat sample. No separation on the back legs and the back is nearly a straight line. It doesn't seem very good at separating like colored objects.

This comment is just confounding. Why would the back leg be separate? It's a cat. Cat's don't have detachable legs.

The selection has the cat's back legs fused together. Same for the dog. Adobe's magic wand could do a better job than this example.