PINGMapper v2.0.0-alpha

[CameronBodine]

PINGMapper v2.0.0-alpha adds to existing functionality from v1.0.0 and many bug fixes. The new features will be documented in a forthcoming manuscript, at which time a production-ready release will be made. Bugs are expected!!! Please report them here.

New Features

Automated Substrate Classification
Neural network models trained with Segmentation Gym have been incorporated into PINGMapper. The models will perform a pixel-wise prediction across 6 different substrate classes [Fines - Rippled, Fines - Flat, Cobble - Boulder, Hard Bottom, Wood, Other]. Plots of the predictions can be optionally exported. Raster and polygon maps can also be exported and overlayed on the sonar mosaics.

NOTE: Exercise caution when interpreting and using the outputs from the substrate prediction. The models were trained on two river systems in Mississippi. It is currently unknown how well the models will perform on other aquatic systems.

Image Corrections
To correct for the impact of attenuation on the sonar imagery, a new feature called Empirical Gain Normalization (EGN) is now available. This process involves calculating the average pixel intensity for each range bin and dividing the raw backscatter by the associated average.

NOTE: Correcting imagery with EGN does take some time; please be patient.

Use Matplotlib colormaps on sonar mosaics
You can now assign one of matplotlibs many colormaps to sonar mosaics.

Ready to get started?

New PINGMapper Users
Please follow the installation instructions.

Existing PINGMapper Users
Update your current installation by following these instructions.

Then check to make sure everything is running as expected by running the test.

---

This discussion was created from the release PINGMapper v2.0.0-alpha.