Research map / Architectures and efficiency / CNN design
Learning Adaptive Receptive Fields for Deep Image Parsing Network
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 |
|---|---|---|---|
| Replace fixed dilation selection with learned inflation, initialized at dilation 8 | Single-path image parsing network · Helen | F-score: 0.8983 -> 0.9021 | Helped |
| Replace fixed dilation selection with learned inflation, initialized at dilation 18 | Single-path image parsing network · PASCAL VOC 2012 validation | mean IOU (%): 63.860 -> 67.780 | Helped |
| Learn path-specific inflation factors with loss guidance | Bi-path parsing network · Helen | F-score: 0.8964 -> 0.9008 | Helped |
| Learn path-specific inflation factors with loss guidance | Tri-path parsing network · Helen | F-score: 0.8894 -> 0.8983 | Helped |
| Add learned inflation initialized at dilation 6 (ablation) | Single-path image parsing network · Helen | F-score: 0.9001 -> 0.8995 | No clear effect |
| Add learned inflation initialized at dilation 2 (ablation) | Single-path image parsing network · Helen | F-score: 0.8995 -> 0.8964 | Hurt |
| Delay updating learned inflation until after 10,000 iterations (ablation) | Single-path model initialized at dilation 2 · Helen | F-score: 0.8995 -> 0.8952 | Hurt |
| Add learned inflation initialized at dilation 6 (ablation) | Single-path image parsing network · PASCAL VOC 2012 validation | mean IOU (%): 64.040 -> 65.080 | Helped |
| Add learned inflation initialized at dilation 4 (ablation) | Single-path image parsing network · PASCAL VOC 2012 validation | mean IOU (%): 61.310 -> 64.536 | Helped |
| Add learned inflation initialized at dilation 16 (ablation) | Single-path image parsing network · PASCAL VOC 2012 validation | mean IOU (%): 64.190 -> 66.030 | Helped |
| Add learned inflation initialized at dilation 20 (ablation) | Single-path image parsing network · PASCAL VOC 2012 validation | mean IOU (%): 63.393 -> 66.530 | Helped |
| Initialize the trainable receptive field too small (ablation) | Learned-inflation parsing network · PASCAL VOC 2012 validation | mean IOU (%): 64.536; 65.080 | Mixed |
Closest papers by meaning
- Understanding the Effective Receptive Field in Deep Convolutional Neural Networks (NeurIPS 2016 · similarity 0.87)
- Receptive Fields As Experts in Convolutional Neural Architectures (ICML 2024 · similarity 0.87)
- QuARF: Quality-Adaptive Receptive Fields for Degraded Image Perception (AAAI 2025 · similarity 0.85)
- Improving the Effective Receptive Field of Message-Passing Neural Networks (ICML 2025 · similarity 0.84)
- PICNN: A Pathway towards Interpretable Convolutional Neural Networks (AAAI 2024 · similarity 0.83)
- Your Image is My Video: Reshaping the Receptive Field via Image-To-Video Differentiable AutoAugmentation and Fusion (CVPR 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.