Research map / Detection and segmentation / Open-set and few-shot detection
Towards a Unified Compositional Model for Visual Pattern Modeling
Wei Tang, Pei Yu, Jiahuan Zhou, Ying Wu
Read the paperCheck 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 |
|---|---|---|---|
| replace LeNet-5 conv/fc layers with And/Or layers (CompNet) | MNIST digit classifier · MNIST test | test error (%): 0.89% -> 0.80% | Improved |
| use CompNet And/Or layers instead of plain CNN with same architecture | character classifier · ICDAR-03 | accuracy (%): 83.0 -> 85.3 | Improved |
| use CompNet instead of CNN with same architecture | character classifier · IIIT5K | accuracy (%): 87.7 -> 88.9 | Improved |
| use CompNet instead of CNN with same architecture | character classifier · Chars74K | accuracy (%): 83.3 -> 85.3 | Improved |
| use CompNet instead of Maxout network with same architecture | character classifier · ICDAR-03 | accuracy (%): 82.7 -> 85.3 | Improved |
| add Chars74K to the training set | CompNet · ICDAR-03 | accuracy (%): 85.3 -> 88.1 | Improved |
| use CompNet on AlexNet conv5 instead of R-CNN with pool5 features | R-CNN detector · PASCAL VOC 2007 | mAP (%): 47.3 -> 52.0 | Improved |
| use CompNet instead of CNN-DPM on same AlexNet conv5 features | detector · PASCAL VOC 2007 | mAP (%): 46.9 -> 52.0 | Improved |
| swap AlexNet for pretrained VGG16 as feature extractor | CompNet detector · PASCAL VOC 2007 | mAP (%): 52.0 -> 67.4 | Improved |
| use CompNet on conv5 instead of R-CNN with fc6 and fc7 layers | detector · PASCAL VOC 2007 | mAP (%): 52.0 -> 58.5 | Hurt |
Closest papers by meaning
- Complexity of Representation and Inference in Compositional Models with Part Sharing (JMLR 2016 · similarity 0.91)
- Does Data Scaling Lead to Visual Compositional Generalization? (ICML 2025 · similarity 0.88)
- Learning to Compose: Improving Object Centric Learning by Injecting Compositionality (ICLR 2024 · similarity 0.88)
- Not Only Text: Exploring Compositionality of Visual Representations in Vision-Language Models (CVPR 2025 · similarity 0.87)
- Compositional Scene Understanding through Inverse Generative Modeling (ICML 2025 · similarity 0.87)
- Scalable Evaluation and Neural Models for Compositional Generalization (NeurIPS 2025 · 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.
