Rethinking BiSeNet For Real-time Semantic Segmentation

Benchmark Model Rank Results
dichotomous-image-segmentation-on-dis-te1STDC#12max F-Measure: 0.648weighted F-measure: 0.562MAE: 0.090
dichotomous-image-segmentation-on-dis-te2STDC#14max F-Measure: 0.720weighted F-measure: 0.636MAE: 0.092
dichotomous-image-segmentation-on-dis-te3STDC#16max F-Measure: 0.745weighted F-measure: 0.662MAE: 0.090
dichotomous-image-segmentation-on-dis-te4STDC#15max F-Measure: 0.731weighted F-measure: 0.652MAE: 0.102
dichotomous-image-segmentation-on-dis-vdSTDC#15max F-Measure: 0.696weighted F-measure: 0.613MAE: 0.103
real-time-semantic-segmentation-on-cityscapesSTDC2-75#8mIoU: 76.8%Frame (fps): 97.0(1080Ti)
real-time-semantic-segmentation-on-cityscapesSTDC1-75#14mIoU: 75.3%Frame (fps): 126.7
real-time-semantic-segmentation-on-cityscapesSTDC2-50#22mIoU: 73.4%Frame (fps): 188.6
real-time-semantic-segmentation-on-cityscapesSTDC1-50#25mIoU: 71.9%Frame (fps): 250.4(1080Ti)
real-time-semantic-segmentation-on-cityscapes-1STDC2-Seg75#11mIoU: 77%Frame (fps): 97
real-time-semantic-segmentation-on-cityscapes-1STDC1-Seg75#19mIoU: 74.5%Frame (fps): 126.7
semantic-segmentation-on-bdd100k-valSTDC2#16mIoU: 53.8(33.0FPS)
semantic-segmentation-on-bdd100k-valSTDC1#20mIoU: 52.1(45.8FPS)