Kelp forest detection from Sentinel-2 imagery using a 2D-CNN (SepConv-SE, 4 classes, 93.7% test accuracy). Pre-trained TFLite models + 148-scene Cedros Island case study (2015–2025) analysing cloud and tide effects on canopy area.
python machine-learning geospatial remote-sensing classification climate-change kelp-forest sentinel-2 macrocystis-pyrifera northeast-pacific fractional-cover
-
Updated
Apr 30, 2026 - Python