Research map / Detection and segmentation / CNN features
Convolutional Oriented Boundaries: From Image Segmentation to High-Level Tasks
Kevis-Kokitsi Maninis, Jordi Pont-Tuset, Pablo Arbeláez, Luc Van Gool
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 |
|---|---|---|---|
| replace local gradient orientation with trained orientations | COB · contour orientation (8 bins) | area under accuracy curve (%): ours 58.6%, local gradients 41.2% | Improved |
| add globalization on top of trained orientations (ablation) | COB (ResNet50) · VOC val | Fb ODS: 0.737 -> 0.734 | Hurt |
| remove globalization step and compute three scales in one pass | COB · PASCAL Context timing | total time per image (s): 24.37 -> 0.79 | Improved |
| use sparse boundary representation for OWT and UCM | MCG-HED · PASCAL Context timing | OWT/UCM time (s): 11.58 -> 1.63 | Improved |
| use ResNet50 instead of VGGNet as base network | COB · VOC val | Fb ODS: 0.705 -> 0.734 | Improved |
| use even deeper ResNet counterparts (ablation) | COB · VOC val | accuracy | No clear effect |
| use COB hierarchy instead of raw side outputs of the base CNN | ResNet50-Side · VOC val | Fb ODS: 0.676 -> 0.737 | Improved |
| replace Structured Edges with HED contours in MCG | MCG · VOC val | Fb ODS: 0.548 -> 0.691 | Improved |
| add orientation and UCM post-processing to RGB-HHA CNN | ResNet50-RGB-HHA · NYUD-v2 | Fb ODS: 0.745 -> 0.784 | Improved |
| add depth as 4th channel instead of HHA encoding (ablation) | ResNet50 on NYUD-v2 · NYUD-v2 | Fb ODS: 0.745 -> 0.683 | Hurt |
| add HHA depth to RGB input | COB · NYUD-v2 | Fb ODS: 0.778 -> 0.784 | Improved |
| combine COB boundaries with dilated network semantic segmentation | DilatedConv · SBD val semantic boundaries | mean maximal Fb: 69.0 -> 70.7 | Improved |
Closest papers by meaning
- Convolutional Oriented Boundaries (ECCV 2016 · similarity 0.99)
- Learning on the Edge: Investigating Boundary Filters in CNNs (IJCV 2019 · similarity 0.89)
- Semantic Segmentation With Boundary Neural Fields (CVPR 2016 · similarity 0.87)
- Devil Is in the Edges: Learning Semantic Boundaries From Noisy Annotations (CVPR 2019 · similarity 0.87)
- DeepLab: Semantic Image Segmentation with Deep Convolutional Nets, Atrous Convolution, and Fully Connected CRFs (TPAMI 2017 · similarity 0.87)
- Learning Hierarchical Image Segmentation For Recognition and By Recognition (ICLR 2024 · similarity 0.86)
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.
