Semantic Segmentation of Aerial Drone Images
Benchmarked UNet, DeepLabV3+, SegNet, and Swin Transformer on 4000×6000 px drone imagery across 23 semantic classes; UNet achieved best accuracy at 86.4%.
Benchmarked UNet, DeepLabV3+, SegNet, and Swin Transformer on 4000×6000 px drone imagery across 23 semantic classes; UNet achieved best accuracy at 86.4%.