A new deep learning semantic segmentation workflow for land cover classification of oblique ground-based mountain photographs
Coming out of the Mountain Legacy Project’s ongoing AI and spatial science innovations, we recently published on the Python Landscape Classifier (PyLC) in The Canadian Journal of Remote Sensing. The publication is open access to any readers.
PyLC presents a major advancement in the automated classification of ground-based oblique images. PyLC was originally developed by then Master’s student Spencer Rose, who has continued working with the project as the developer for our online database explore.mountainlegacy.ca. Aniket Mahindrakar continued advancing PyLC as part of his own Master’s along with research associate Ben Wright. We also want to thank all the contributors who organized and provided manually annotated training data including Darcy Benham, James Tricker, Sebastian Yerex, Elizabeth Alrich, Kristyn Lang, Larissa Bron, Tanya Taggart-Hodge, Julie Fortin, and Maya Frederickson.
All training data for PyLC is now available in an open access repository as the MountainScape Segmentation Dataset (MS2D). The code for running PyLC and training new models is available through GitHub. PyLC also operates within the Mountain Image Analysis Suite (MIASuite), an open source plugin for QGIS.
Image 1. Example output for historical and repeat images from the MountainScape Segmentation Dataset classified using the Python Landscape Classifier. The bottom panels show manually segmented masks.
