Research map / Detection and segmentation / CNN features
Semantic Object Parsing With Local-Global Long Short-Term Memory
Xiaodan Liang, Xiaohui Shen, Donglai Xiang, Jiashi Feng, Liang Lin, Shuicheng Yan
Check your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 stacked LG-LSTM layers to the convolutional network (ablation) | VGG-16 parsing network · Horse parsing | IoU: 64.13 -> 68.73; 4.6% improvement | Improved |
| Replace LG-LSTM layers with five extra convolutional layers (ablation) | VGG-16 parsing network · Horse parsing | IoU: 68.73 -> 65.95; 2.78% | Hurt |
| Replace LG-LSTM layers with five extra convolutional layers (ablation) | VGG-16 parsing network · Cow parsing | IoU: 62.79 -> 57.93; 4.86% | Hurt |
| Use eight local connections instead of two (ablation) | LG-LSTM · Cow parsing | IoU: 58.60 -> 62.79; 4.19% | Improved |
| Use eight local connections instead of four (ablation) | LG-LSTM · Cow parsing | IoU: 59.85 -> 62.79; 2.94% | Improved |
| Add global hidden cells to the LSTM inputs (ablation) | LG-LSTM · Horse parsing | IoU: 67.46 -> 68.73; 1.27% | Improved |
| Add global hidden cells to the LSTM inputs (ablation) | LG-LSTM · Cow parsing | IoU: 60.98 -> 62.79; 1.81% | Improved |
| Train with 10,000 additional images | LG-LSTM · ATR | Avg. F-1 score: 80.97 -> 84.12 | Improved |
| Train with 10,000 additional images | Co-CNN · ATR | Avg. F-1 score: 76.95 -> 80.14 | Improved |
| Increase the LG-LSTM stack beyond five layers | LG-LSTM · Object and human parsing datasets | performance | No clear effect |
Closest papers by meaning
- Semantic Object Parsing with Graph LSTM (ECCV 2016 · similarity 0.92)
- Deep Structured Scene Parsing by Learning With Image Descriptions (CVPR 2016 · similarity 0.87)
- MMLongBench: Benchmarking Long-Context Vision-Language Models Effectively and Thoroughly (NeurIPS 2025 · similarity 0.85)
- Grounding Everything: Emerging Localization Properties in Vision-Language Transformers (CVPR 2024 · similarity 0.85)
- Vision-LSTM: xLSTM as Generic Vision Backbone (ICLR 2025 · similarity 0.85)
- GOAL: Global-local Object Alignment Learning (CVPR 2025 · similarity 0.85)
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.
