Post Hoc Denoising

Detection in DeepSqueak v3 generally produces fewer false positives than v2, so a default post-hoc denoising network has not been included. A new post-hoc denoising network may be trained for recordings with particularly bad noise issues.

False positives may be automatically identified and rejected using a post-hoc neural network.

To use the post hoc denoiser:

1. Select "Tools > Automatic Review > Post Hoc Denoising"

2. In the list box, select the detection files to denoise. All noise events found will be classified as "Noise" and rejected.

To train a post hoc denoiser:

1. Select "Tools > Network Training > Train Post Hoc Denoiser"

2. In the list box, select the detection files to to use for training

   • Calls labeled as "Noise" are used as negative training sample

     • Tip: Use the "add custom labels" tool to hand label noise, more variety is better
     • Tip: Make sure all types of "true" calls are well represented in the training set

   • Accepted calls not labeled as "Noise" are used as positive training samples.

3. When training is finished, a save dialog box will appear to save the denoising network