Enhancing crop segmentation in satellite image time-series with transformer networks

Benchmark Model Rank Results
semantic-segmentation-on-lombardia-sentinel-2UNet3D#1Overall Accuracy: 80.77
semantic-segmentation-on-lombardia-sentinel-2Swin UNETR#2Overall Accuracy: 79.64
semantic-segmentation-on-lombardia-sentinel-23D FPN with NDVI Loss#3Overall Accuracy: 77.23
semantic-segmentation-on-lombardia-sentinel-2DeepLabv3 3D#4Overall Accuracy: 74.51
unet-segmentation-on-munich-sentinel2-crop-1Swin UNETR#1Overall Accuracy: 95.26
unet-segmentation-on-munich-sentinel2-crop-1UNet3D#2Overall Accuracy: 94.73
unet-segmentation-on-munich-sentinel2-crop-1DeepLabv3 3D#4Overall Accuracy: 85.98