Research map / Detection and segmentation / Weakly supervised detection and segmentation
Deep Structured Scene Parsing by Learning With Image Descriptions
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 |
|---|---|---|---|
| Train the CNN-RNN model using image descriptions and weak supervision | CNN-RNN scene parsing · PASCAL VOC 2012 val | IoU: 30.3% -> 35.2% | Helped |
| Add a trainable RNN rather than fixing it at random initialization (ablation) | CNN-RNN scene parsing · PASCAL VOC 2012 val | IoU: 33.5% -> 35.2% | Helped |
| Use fusion training with 280 strong samples alongside weakly annotated samples | CNN-RNN scene labeling · PASCAL VOC 2012 val | IoU: 35.2% -> 45.3% | Helped |
| Use fusion rather than waterfall training with the same 280 strong samples (ablation) | CNN-RNN scene labeling · PASCAL VOC 2012 val | IoU: 35.1% -> 45.3% | Helped |
| Use 1464 rather than 280 strongly annotated samples with fusion training | CNN-RNN scene labeling · PASCAL VOC 2012 val | IoU: 45.3% -> 52.9% | Helped |
| Add joint end-to-end CNN-RNN learning rather than fixing part of the CNN (ablation) | CNN-RNN structured scene parsing · PASCAL VOC 2012 val | structure accuracy: 57.0% -> 67.4% | Helped |
| Add joint end-to-end CNN-RNN learning rather than fixing part of the CNN (ablation) | CNN-RNN structured scene parsing · PASCAL VOC 2012 val | relation accuracy: 49.0% -> 62.8% | Helped |
| Train the RNN instead of keeping its parameters randomly initialized (ablation) | CNN-RNN structured scene parsing · PASCAL VOC 2012 val | structure accuracy: 40.8% -> 67.4% | Helped |
| Train the RNN instead of keeping its parameters randomly initialized (ablation) | CNN-RNN structured scene parsing · PASCAL VOC 2012 val | relation accuracy: 31.4% -> 62.8% | Helped |
| Train CNN and RNN jointly rather than in separate stages (ablation) | CNN-RNN structured scene parsing · PASCAL VOC 2012 val | structure accuracy: 60.8% -> 67.4% | Helped |
| Train CNN and RNN jointly rather than in separate stages (ablation) | CNN-RNN structured scene parsing · PASCAL VOC 2012 val | relation accuracy: 54.2% -> 62.8% | Helped |
Closest papers by meaning
- Weakly Supervised Scene Parsing with Point-Based Distance Metric Learning (AAAI 2019 · similarity 0.90)
- Weakly Supervised Visual Semantic Parsing (CVPR 2020 · similarity 0.89)
- Multi-Path Feedback Recurrent Neural Networks for Scene Parsing (AAAI 2017 · similarity 0.88)
- Unsupervised Vision-Language Parsing: Seamlessly Bridging Visual Scene Graphs With Language Structures via Dependency Relationships (CVPR 2022 · similarity 0.88)
- Reading Scene Text in Deep Convolutional Sequences (AAAI 2016 · similarity 0.87)
- Aligning Forest and Trees in Images & Long Captions for Visually Grounded Understanding (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.