Research map / Label-efficient and robust learning / Recognition and long-tail
DAG-Recurrent Neural Networks For Scene Labeling
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 DAG-RNN(8) contextual dependencies to CNN-65 | CNN-65 · SiftFlow | Global accuracy: 76.1% -> 81.1% | Helped |
| add DAG-RNN(8) contextual dependencies to CNN-65 | CNN-65 · CamVid | Class accuracy: 53.2% -> 68.3% | Helped |
| add DAG-RNN(8) contextual dependencies to VGG-conv5 | VGG-conv5 · SiftFlow | Class accuracy: 48.8% -> 55.7%; 6.9% | Helped |
| add DAG-RNN(8) contextual dependencies to VGG-conv5 | VGG-conv5 · CamVid | Global accuracy: 91.0% -> 91.6% | Helped |
| add DAG-RNN(8) contextual dependencies to VGG-conv5 | VGG-conv5 · Barcelona | Class accuracy: 21.1% -> 24.6% | Helped |
| use 8-neighborhood instead of 4-neighborhood DAG-RNN (ablation) | CNN-65-DAG-RNN · SiftFlow | Class accuracy: 0.6% / 5.6% | Helped |
| use 8-neighborhood instead of 4-neighborhood DAG-RNN (ablation) | CNN-65-DAG-RNN · CamVid | Class accuracy: 0.5% / 2.0% | Helped |
| use 8-neighborhood instead of 4-neighborhood DAG-RNN (ablation) | CNN-65-DAG-RNN · Barcelona | Class accuracy: 1.1% / 1.6% | Helped |
| add class weighting for rare classes (ablation) | CNN-65 · SiftFlow | Class accuracy: 32.5% -> 37.0% | Helped |
| jointly optimize CNN-65 and DAG-RNN(8), rather than optimize them independently (ablation) | CNN-65-DAG-RNN(8) · SiftFlow | Class accuracy: 80.1% and 42.7% | Helped |
| increase hidden-layer capacity beyond the selected size (ablation) | DAG-RNN · SiftFlow | accuracy | No clear effect |
Closest papers by meaning
- CNN-RNN: A Unified Framework for Multi-Label Image Classification (CVPR 2016 · similarity 0.87)
- Multi-Path Feedback Recurrent Neural Networks for Scene Parsing (AAAI 2017 · similarity 0.86)
- Semantic Regularisation for Recurrent Image Annotation (CVPR 2017 · similarity 0.86)
- Order-Free RNN With Visual Attention for Multi-Label Classification (AAAI 2018 · similarity 0.85)
- GreedyViG: Dynamic Axial Graph Construction for Efficient Vision GNNs (CVPR 2024 · similarity 0.84)
- Adaptive Self-training Framework for Fine-grained Scene Graph Generation (ICLR 2024 · similarity 0.84)
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.