Training detection network

[livelovelarynx]

Working on training a detection network in DeepSqueak. I've been using the detection files themselves to create images for training. The detector still struggles to ID complex calls (especially harmonics). Any tips on creating training images in DeepSqueak? Would you suggest using a different software (ex: Raven Lite) to create the training images?

I've been using the following commands within DeepSqueak to create images and then train the detector:

1. Tools->Network training->Create detection network training images.
2. Tools->Network training->Train detection network.

Please let me know if you have any advice. Thank you!