1. JSR381 Standard Compliant
As of 2.1.0 release Deep Netts is fully compatible with final release of JSR381 standard Java API for visual recognition
Learn more about JSR381 here
Full examples for visual recognition and machine learning using JSR381 are available on GitHub
2. Better user experience and debugging tools
It fixes a number of small bug in GUI, and brings user-friendly tool for testing and debugging image recognition, which enables you to easily test trained neural network with selected images and better understand what and why is not working.
3. Improved performance for all types of networks
Optimized multi-threaded calculation for fully connected and output layers brings significant performance improvement for all types of neural networks.