Semantic Segmentation On Stanford2D3D 1

评估指标

mAcc

mIoU

评测结果

各个模型在此基准测试上的表现结果

			Paper Title	Repository
SFSS-MMSI (RGB+HHA)	70.68	60.6%	Single Frame Semantic Segmentation Using Multi-Modal Spherical Images
SFSS-MMSI (RGB+Depth+Normal)	69.03	59.43%	Single Frame Semantic Segmentation Using Multi-Modal Spherical Images
SFSS-MMSI (RGB+Normal)	68.79	58.24%	Single Frame Semantic Segmentation Using Multi-Modal Spherical Images
SGAT4PASS（RGB only, Fold 1）	-	56.4%	SGAT4PASS: Spherical Geometry-Aware Transformer for PAnoramic Semantic Segmentation
SFSS-MMSI (RGB+Depth)	68.57	55.49%	Single Frame Semantic Segmentation Using Multi-Modal Spherical Images
SGAT4PASS（RGB only, 3 Fold AVG）	-	55.3%	SGAT4PASS: Spherical Geometry-Aware Transformer for PAnoramic Semantic Segmentation
Trans4PASS+ (Supervised + Small + MS)	-	54.0%	Behind Every Domain There is a Shift: Adapting Distortion-aware Vision Transformers for Panoramic Semantic Segmentation
Trans4PASS (Supervised + Small + MS)	-	53.0%	Bending Reality: Distortion-aware Transformers for Adapting to Panoramic Semantic Segmentation
SFSS-MMSI (RGB Only)	63.96	52.87%	Single Frame Semantic Segmentation Using Multi-Modal Spherical Images
Trans4PASS+ (UDA + MPA + MS)	-	52.3%	Behind Every Domain There is a Shift: Adapting Distortion-aware Vision Transformers for Panoramic Semantic Segmentation
CBFC	65.6	52.2%	Complementary Bi-directional Feature Compression for Indoor 360° Semantic Segmentation with Self-distillation	-
Trans4PASS (Supervised + Small)	-	52.1%	Bending Reality: Distortion-aware Transformers for Adapting to Panoramic Semantic Segmentation
HoHoNet (ResNet-101)	65.0	52.0%	HoHoNet: 360 Indoor Holistic Understanding with Latent Horizontal Features
Trans4PASS (UDA + MPA + MS)	-	51.2%	Bending Reality: Distortion-aware Transformers for Adapting to Panoramic Semantic Segmentation
Trans4PASS (UDA + MPA)	-	50.8%	Bending Reality: Distortion-aware Transformers for Adapting to Panoramic Semantic Segmentation
PanoFormer	64.5	48.9%	PanoFormer: Panorama Transformer for Indoor 360 Depth Estimation
Trans4PASS (UDA + Source Only)	-	48.1%	Bending Reality: Distortion-aware Transformers for Adapting to Panoramic Semantic Segmentation
FreDSNet	63.1	46.1%	FreDSNet: Joint Monocular Depth and Semantic Segmentation with Fast Fourier Convolutions
Tangent (ResNet-101)	65.2	45.6%	Tangent Images for Mitigating Spherical Distortion
SWSCNN	-	43.4%	Spin-Weighted Spherical CNNs

0 of 25 row(s) selected.