Research map / Detection and segmentation / CNN features
UCTNet: Uncertainty-Aware Cross-Modal Transformer Network for Indoor RGB-D Semantic Segmentation
Xiaowen Ying, Mooi Choo Chuah
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add depth map as 4th input channel | Swin-S RGB baseline · NYUv2 | mIoU (%): 52.2 -> 52.4 | Improved |
| add a separate depth encoder with element-wise add fusion (ablation) | RGB-only Swin-S baseline · NYUv2 | mIoU (%): 52.2 -> 54.3 | Improved |
| use our fusion module instead of element-wise add (ablation) | dual transformer encoders · NYUv2 | mIoU (%): 54.3 -> 55.3 | Improved |
| use our fusion module instead of SE-Add (ablation) | dual transformer encoders · NYUv2 | mIoU (%): 54.6 -> 55.3 | Improved |
| use our fusion module instead of SA-Gate (ablation) | dual transformer encoders · NYUv2 | mIoU (%): 53.8 -> 55.3 | Improved |
| replace standard depth encoder with UASA cut-off (ablation) | UCTNet with our fusion · NYUv2 | mIoU (%): 55.3 -> 56.8 | Improved |
| replace standard depth encoder with UASA suppression (ablation) | UCTNet with our fusion · NYUv2 | mIoU (%): 55.3 -> 57.6 | Improved |
| use UASA suppression instead of cut-off (ablation) | UCTNet depth encoder · NYUv2 | mIoU (%): 56.8 -> 57.6 | Improved |
| raise temperature T from 5 to 15 (ablation) | UASA suppression · NYUv2 | mIoU (%): 57.2 -> 57.6 | Improved |
| raise temperature T from 15 to 20 (ablation) | UASA suppression · NYUv2 | mIoU (%): 57.6 -> 56.8 | Hurt |
| raise temperature T from 5 to 10 (ablation) | UASA suppression · NYUv2 | mIoU (%): 57.2 -> 57.2 | No clear effect |
| UCTNet versus previous best without multi-scale testing | UCTNet · NYUv2 | mIoU (%): 57.6 | Improved |
Closest papers by meaning
- Locality-Sensitive Deconvolution Networks With Gated Fusion for RGB-D Indoor Semantic Segmentation (CVPR 2017 · similarity 0.91)
- Learning Common and Specific Features for RGB-D Semantic Segmentation with Deconvolutional Networks (ECCV 2016 · similarity 0.89)
- Re-coding for Uncertainties: Edge-awareness Semantic Concordance for Resilient Event-RGB Segmentation (NeurIPS 2025 · similarity 0.87)
- PDDM: Pseudo Depth Diffusion Model for RGB-PD Semantic Segmentation Based in Complex Indoor Scenes (AAAI 2025 · similarity 0.86)
- DFormer: Rethinking RGBD Representation Learning for Semantic Segmentation (ICLR 2024 · similarity 0.86)
- DFormerv2: Geometry Self-Attention for RGBD Semantic Segmentation (CVPR 2025 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
