Research map / Detection and segmentation / CNN features
Cascaded Feature Network for Semantic Segmentation of RGB-D Images
Di Lin, Guangyong Chen, Daniel Cohen-Or, Pheng-Ann Heng, Hui Huang
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 |
|---|---|---|---|
| use 3 depth-based branches instead of a single branch (ablation) | CFN (VGG-16) · NYUDv2 validation | mean IoU (%): 31.2 -> 35.5 | Improved |
| use 4 or 5 branches instead of 3 (larger super-pixels) (ablation) | CFN (VGG-16) · NYUDv2 validation | mean IoU (%): 35.5 -> 35.0 -> 34.6 | Hurt |
| add CaRF to single-branch network (ablation) | VGG-FCN · NYUDv2 test | mean IoU (%): 31.8 -> 32.0 | Improved |
| add CaRF to 3-branch network (ablation) | CFN (VGG-16) · NYUDv2 test | mean IoU (%): 31.7 -> 36.3 | Improved |
| add multiple branches without CaRF (ablation) | VGG-FCN · NYUDv2 test | mean IoU (%): 31.8 -> 31.7 | No clear effect |
| use CaRF instead of regular receptive fields (pooling kernels 3,5,7) (ablation) | 3-branch CFN · NYUDv2 test | mean IoU (%): 33.8 -> 36.3 | Improved |
| add regular receptive fields to 3-branch network (ablation) | 3-branch CFN · NYUDv2 test | mean IoU (%): 31.7 -> 33.8 | Improved |
| add cascaded links between branches (ablation) | separate-branch CFN · NYUDv2 test | mean IoU (%): 33.0 -> 36.3 | Improved |
| use separate contextual information per scene-resolution instead of shared (ablation) | combined-branch CFN · NYUDv2 test | mean IoU (%): 34.0 -> 36.3 | Improved |
| use deeper RefineNet-152 instead of VGG-16 in CFN | CFN · NYUDv2 test | mean IoU (%): 41.7 -> 47.7 | Improved |
| use deeper RefineNet-152 instead of VGG-16 in CFN | CFN · SUN-RGBD test | mean IoU (%): 42.5 -> 48.1 | Improved |
| add CFN to RefineNet-152 baseline (RGB-D) | RefineNet-152 · NYUDv2 test | mean IoU (%): 47.0 -> 47.7 | Improved |
Closest papers by meaning
- Zig-Zag Network for Semantic Segmentation of RGB-D Images (TPAMI 2019 · similarity 0.92)
- Semantic Segmentation With Boundary Neural Fields (CVPR 2016 · similarity 0.89)
- Fully Convolutional Neural Networks with Full-Scale-Features for Semantic Segmentation (AAAI 2017 · similarity 0.89)
- Context Encoding for Semantic Segmentation (CVPR 2018 · similarity 0.89)
- Depth-aware CNN for RGB-D Segmentation (ECCV 2018 · similarity 0.89)
- Unsupervised Semantic Segmentation Through Depth-Guided Feature Correlation and Sampling (CVPR 2024 · similarity 0.88)
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.
