Fetching the paper…

Per-Pixel Classification is Not All You Need for Semantic Segmentation · Around