Speed improvement (10 FPS -> 40 FPS)

We greatly improved speed of image reading and preprocessing. Now it takes only 5% of total time instead of 77% earlier. Speed for 8 convolution version of device increased from ~10 FPS up to ~ 40 FPS.

Pre-trained models

v1.0

Initial release