Research map / Detection and segmentation / CNN features
ZigZagNet: Fusing Top-Down and Bottom-Up Context for Object Segmentation
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 bottom-up context exchange and dense pathways (ablation) | ZigZagNet encoder-decoder variant · PASCAL VOC 2012 val | mIoU: 82.5 -> 86.0 | Helped |
| Add feature-map fusion (ablation) | ZigZagNet · PASCAL VOC 2012 val | mIoU: 83.6 -> 86.0 | Helped |
| Add context exchange between top-down and bottom-up networks (ablation) | ZigZagNet · PASCAL VOC 2012 val | mIoU: 84.2 -> 86.0 | Helped |
| Add dense pathways with region context encoding (ablation) | ZigZagNet · PASCAL VOC 2012 val | mIoU: 84.9 -> 86.0 | Helped |
| Use RCE rather than global context encoding (ablation) | ZigZagNet · PASCAL VOC 2012 val | mIoU: 82.7 -> 86.0; 83.0 -> 86.0 | Helped |
| Use RCE rather than large-kernel context encoding (ablation) | ZigZagNet · PASCAL VOC 2012 val | mIoU: 83.4 -> 86.0 | Helped |
| Use RCE rather than spatial pyramid pooling (ablation) | ZigZagNet · PASCAL VOC 2012 val | mIoU: 84.1 -> 86.0; 84.7 -> 86.0 | Helped |
| Use multiplication rather than addition to combine context feature maps (ablation) | ZigZagNet · PASCAL VOC 2012 val | mIoU: +0.8 score | Helped |
| Add ZigZagNet feature context | FCIS, Deformable RCNN, and Mask RCNN · COCO val | mask AP: 29.2 →32.2; 31.7 →33.4; 36.1 →38.2; 37.9 →39.8; 37.5 →39.5; 39.7 →40.8 | Helped |
| Use ZigZagNet for bidirectional multi-scale context propagation | semantic segmentation · PASCAL VOC 2012 val | mIoU: 85.8 -> 86.0 | Helped |
Closest papers by meaning
- SegNeXt: Rethinking Convolutional Attention Design for Semantic Segmentation (NeurIPS 2022 · similarity 0.85)
- Decoupling Zero-Shot Semantic Segmentation (CVPR 2022 · similarity 0.84)
- Scaling up Image Segmentation across Data and Tasks (CVPR 2025 · similarity 0.84)
- LangHOPS: Language Grounded Hierarchical Open-Vocabulary Part Segmentation (NeurIPS 2025 · similarity 0.83)
- SegMAN: Omni-scale Context Modeling with State Space Models and Local Attention for Semantic Segmentation (CVPR 2025 · similarity 0.83)
- CSL: Class-Agnostic Structure-Constrained Learning for Segmentation Including the Unseen (AAAI 2024 · similarity 0.83)
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.