Unsupervised Semantic Segmentation by Contrasting Object Mask Proposals

Benchmark Model Rank Results
unsupervised-semantic-segmentation-on-6MaskContrast (+Saliency map)#3mIoU (test): 24.2mIoU (val): 24.6
unsupervised-semantic-segmentation-on-coco-8MaskContrast (ResNet-50)#4mIoU: 3.7Pixel Accuracy: 8.8
unsupervised-semantic-segmentation-on-pascal-1MaskContrast (Saliency)#8Clustering [mIoU]: 44.2Linear Classifier [mIoU]: 63.9
unsupervised-semantic-segmentation-on-pascal-1MaskContrast#11Clustering [mIoU]: 35.0Linear Classifier [mIoU]: 58.4