Semantic Segmentation On Nyu Depth V2

评估指标

Mean IoU

评测结果

各个模型在此基准测试上的表现结果

		Paper Title	Repository
OmniVec2	63.6	OmniVec2 - A Novel Transformer based Network for Large Scale Multimodal and Multitask Learning	-
DiffusionMMS (DAT++-S)	61.5	Diffusion-based RGB-D Semantic Segmentation with Deformable Attention Transformer	-
GeminiFusion (Swin-Large)	60.9	GeminiFusion: Efficient Pixel-wise Multimodal Fusion for Vision Transformer
OmniVec	60.8	OmniVec: Learning robust representations with cross modal sharing	-
GeminiFusion (Swin-Large)	60.2	GeminiFusion: Efficient Pixel-wise Multimodal Fusion for Vision Transformer
DPLNet	59.3	Efficient Multimodal Semantic Segmentation via Dual-Prompt Learning
EMSANet (2x ResNet-34 NBt1D, PanopticNDT version, finetuned)	59.02	PanopticNDT: Efficient and Robust Panoptic Mapping
SwinMTL	58.14%	SwinMTL: A Shared Architecture for Simultaneous Depth Estimation and Semantic Segmentation from Monocular Camera Images
PolyMaX(ConvNeXt-L)	58.08%	PolyMaX: General Dense Prediction with Mask Transformer
HSPFormer(PVT v2-B4)	57.8%	HSPFormer: Hierarchical Spatial Perception Transformer for Semantic Segmentation	-
GeminiFusion (MiT-B5)	57.7	GeminiFusion: Efficient Pixel-wise Multimodal Fusion for Vision Transformer
DFormer-L	57.2%	DFormer: Rethinking RGBD Representation Learning for Semantic Segmentation
CMNeXt (B4)	56.9%	Delivering Arbitrary-Modal Semantic Segmentation
CMX (B5)	56.9%	CMX: Cross-Modal Fusion for RGB-X Semantic Segmentation with Transformers
GeminiFusion (MiT-B3)	56.8	GeminiFusion: Efficient Pixel-wise Multimodal Fusion for Vision Transformer
OMNIVORE (Swin-L, finetuned)	56.8%	Omnivore: A Single Model for Many Visual Modalities
CMX (B4)	56.3%	CMX: Cross-Modal Fusion for RGB-X Semantic Segmentation with Transformers
MultiMAE (ViT-B)	56.0%	MultiMAE: Multi-modal Multi-task Masked Autoencoders
SMMCL (SegNeXt-B)	55.8%	Understanding Dark Scenes by Contrasting Multi-Modal Observations
DFormer-B	55.6%	DFormer: Rethinking RGBD Representation Learning for Semantic Segmentation

0 of 116 row(s) selected.