RGBD20K: A Large-Scale Benchmark for RGB-D Semantic Segmentation
The benchmark pairs 20,000 RGB-D images with 160 fine-grained segmentation classes.
The authors position RGBD20K as a larger and more semantically broad alternative to NYUv2 and SUN RGB-D. They say the labels were re-evaluated and corrected to reduce annotation noise. The paper also introduces a score-purified fusion method that reports state-of-the-art results across the evaluated benchmarks. HF Daily Papers' note
The authors position RGBD20K as a larger and more semantically broad alternative to NYUv2 and SUN RGB-D. They say the labels were re-evaluated and corrected to reduce annotation noise. The paper also introduces a score-purified fusion method that reports state-of-the-art results across the evaluated benchmarks. HF Daily Papers' note
score 5