Research map / Detection and segmentation / CNN features
Recurrent Scene Parsing With Perspective Understanding in the Loop
Read the paperCheck 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 depth-aware pooling and two recurrent refinement iterations (ablation) | segmentation model · NYU-depth-v2 | IoU: 0.406 -> 0.431 | Helped |
| add depth-aware pooling and two recurrent refinement iterations (ablation) | segmentation model · SUN-RGBD | IoU: 0.402 -> 0.443 | Helped |
| add depth-aware pooling and two recurrent refinement iterations (ablation) | segmentation model · Stanford-2D-3D | IoU: 0.644 -> 0.760 | Helped |
| add depth-aware pooling and two recurrent refinement iterations (ablation) | segmentation model · Cityscapes | IoU: 0.738 -> 0.776 | Helped |
| average multiple dilation branches rather than use the baseline (ablation) | Cityscapes segmentation model · Cityscapes | IoU: 0.738 -> 0.747 | Helped |
| use ground-truth depth to gate pooling rather than average branches (ablation) | Cityscapes segmentation model · Cityscapes | IoU: 0.747 -> 0.748 | Helped |
| untie weights across scale-specific branches (ablation) | ground-truth-depth gated model · Cityscapes | IoU: 0.748 -> 0.753 | Helped |
| learn the gating signal without depth supervision (ablation) | untied gated model · Cityscapes | IoU: 0.738 -> 0.754 | Helped |
| use predicted depth instead of ground-truth depth for gating (ablation) | untied gated model · Cityscapes | IoU: 0.753 -> 0.759 | Helped |
| apply Gaussian smoothing to predicted depth (ablation) | monocular depth predictor · NYU-depth-v2 | δ<1.25: 0.809 -> 0.816 | Helped |
| use test-time augmentation with flipped and rescaled images (ablation) | loop2 segmentation model · Cityscapes | IoU: 0.776 -> 0.791 | Helped |
| use a second recurrent iteration rather than stopping after loop1 with predicted depth (ablation) | recurrent segmentation model · Cityscapes | IoU: 0.772 -> 0.776 | Helped |
Closest papers by meaning
- SDC-Depth: Semantic Divide-and-Conquer Network for Monocular Depth Estimation (CVPR 2020 · similarity 0.89)
- Unsupervised Semantic Segmentation Through Depth-Guided Feature Correlation and Sampling (CVPR 2024 · similarity 0.89)
- Towards Scene Understanding: Unsupervised Monocular Depth Estimation With Semantic-Aware Representation (CVPR 2019 · similarity 0.89)
- On the Robustness of Language Guidance for Low-Level Vision Tasks: Findings from Depth Estimation (CVPR 2024 · similarity 0.86)
- ResCLIP: Residual Attention for Training-free Dense Vision-language Inference (CVPR 2025 · similarity 0.86)
- RefineAny3D: Depth Refinement as Semantic Alignment for Monocular 3D Detection (NeurIPS 2026 · 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.