For what I was using it for - locating fruit on a plant - detectnet works pretty well (it runs at 7fps on a TX2 at 1280x960). It's a modified GoogLeNet with som

For what I was using it for - locating fruit on a plant - detectnet works pretty well (it runs at 7fps on a TX2 at 1280x960). It's a modified GoogLeNet with some custom layer on the end to do the bounding boxes. DIGITS itself is just a management console. That said, you can train Tensorflow models using DIGITS, so I assume that DIGITS handles the data IO for you: https://github.com/NVIDIA/DIGITS/blob/master/docs/GettingStartedTensorflow.md https://github.com/NVIDIA/DIGITS/blob/master/docs/GettingSta... It looks like you just define a wrapper around your model so that DIGITS knows how to run it and evaluate the loss, but otherwise your model could be anything.