Research map / Architectures and efficiency / CNN design
Densely Connected Multi-Dilated Convolutional Networks for Dense Prediction Tasks
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 |
|---|---|---|---|
| Use D3Net-L instead of HRNetV2-W48 as the backbone in OCRNet | OCRNet on Cityscapes val · Cityscapes val | mIoU: 80.7 -> 81.2 | Helped |
| Use D3Net-L instead of HRNetV2-W48 as the FCN backbone | FCN on Cityscapes test · Cityscapes test | mIoU: 80.4 -> 80.8 | Helped |
| Use D3Net for audio source separation instead of the strongest listed baseline | MUSDB18 source separation · MUSDB18 | average SDR (dB): 5.82 -> 6.01 | Helped |
| Add multidilated convolution rather than using standard convolution (ablation) | D3Net-S on Cityscapes · Cityscapes val | mIoU: 62.3 -> 65.1 | Helped |
| Add multidilated convolution rather than standard dilated convolution (ablation) | D3Net-S on Cityscapes · Cityscapes val | mIoU: 57.9 -> 65.1 | Helped |
| Replace multidilated convolution with standard dilation (ablation) | D3Net-S on Cityscapes · Cityscapes val | mIoU: 62.3 -> 57.9 | Hurt |
| Use nested D3 blocks rather than standard dense blocks (ablation) | D3Net-S on Cityscapes · Cityscapes val | mIoU: 57.6 -> 65.1 | Helped |
| Use nested D3 blocks rather than standard dense blocks with a similar layer count (ablation) | D3Net-S on Cityscapes · Cityscapes val | mIoU: 54.9 -> 65.1 | Helped |
| Add multidilated convolution rather than using standard convolution (ablation) | D3Net on MUSDB18 · MUSDB18 | average SDR (dB): 5.60 -> 6.01 | Helped |
| Add multidilated convolution rather than standard dilated convolution (ablation) | D3Net on MUSDB18 · MUSDB18 | average SDR (dB): 5.86 -> 6.01 | Helped |
Closest papers by meaning
- Densely Connected Convolutional Networks (CVPR 2017 · similarity 0.89)
- Residual Dense Network for Image Super-Resolution (CVPR 2018 · similarity 0.86)
- RFC-Net: Learning High Resolution Global Features for Medical Image Segmentation on a Computational Budget (Student Abstract) (AAAI 2023 · similarity 0.86)
- DCNAS: Densely Connected Neural Architecture Search for Semantic Image Segmentation (CVPR 2021 · similarity 0.85)
- Revisiting Dilated Convolution: A Simple Approach for Weakly- and Semi-Supervised Semantic Segmentation (CVPR 2018 · similarity 0.85)
- DenseMLLM: Standard Multimodal LLMs for Dense Prediction (ICML 2026 · 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.