HyperSeg: Patch-wise Hypernetwork for Real-time Semantic Segmentation

Benchmark Model Rank Results
dichotomous-image-segmentation-on-dis-te1HySM#7max F-Measure: 0.695weighted F-measure: 0.597MAE: 0.082
dichotomous-image-segmentation-on-dis-te2HySM#7max F-Measure: 0.759weighted F-measure: 0.667MAE: 0.085
dichotomous-image-segmentation-on-dis-te3HySM#8max F-Measure: 0.792weighted F-measure: 0.701MAE: 0.079
dichotomous-image-segmentation-on-dis-te4HySM#8max F-Measure: 0.782weighted F-measure: 0.693MAE: 0.091
dichotomous-image-segmentation-on-dis-vdHySM#10max F-Measure: 0.734weighted F-measure: 0.640MAE: 0.096
real-time-semantic-segmentation-on-camvidHyperSeg-L#7mIoU: 79.1Frame (fps): 16.6Time (ms): 60.2
real-time-semantic-segmentation-on-camvidHyperSeg-S#9mIoU: 78.4Frame (fps): 38.0Time (ms): 26.3
real-time-semantic-segmentation-on-cityscapesHyperSeg-M#11mIoU: 75.8%Frame (fps): 36.9Time (ms): 27.1
semantic-segmentation-on-pascal-voc-2012-valHyperSeg-L#7mIoU: 80.61%