PLOP: Learning without Forgetting for Continual Semantic Segmentation

Open paper
Benchmark Model Rank Results
disjoint-10-1-on-pascal-voc-2012PLOP#4mIoU: 8.4
disjoint-15-1-on-pascal-voc-2012PLOP#5mIoU: 46.5
disjoint-15-5-on-pascal-voc-2012PLOP#6Mean IoU: 64.3
overlapped-10-1-on-pascal-voc-2012PLOP#9mIoU: 30.45
overlapped-100-10-on-ade20kPLOP#5Mean IoU (test): 31.59
overlapped-100-10-on-ade20kMiB#6Mean IoU (test): 29.24
overlapped-100-5-on-ade20kPLOP#6mIoU: 28.75
overlapped-100-5-on-ade20kMiB#7mIoU: 25.96
overlapped-100-50-on-ade20kPLOP#6mIoU: 32.94
overlapped-100-50-on-ade20kMiB#7mIoU: 32.79
overlapped-15-1-on-pascal-voc-2012PLOP#9mIoU: 54.64
overlapped-15-1-on-pascal-voc-2012MiB#10mIoU: 29.29
overlapped-15-5-on-pascal-voc-2012PLOP#8Mean IoU (val): 70.09
overlapped-15-5-on-pascal-voc-2012MiB#9Mean IoU (val): 70.08
overlapped-50-50-on-ade20kPLOP#4mIoU: 30.4
overlapped-50-50-on-ade20kMiB#7mIoU: 29.31

Rank counts only results with a code link.