Identify objects in images using a third-generation deep residual network.
Also includes fix to support PNG files
First release of the model