Research map / Image generation / Autoregressive and latent generation
Reliable and Efficient Image Cropping: A Grid Anchor Based Approach
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 |
|---|---|---|---|
| Model both the region of interest and the region of discard instead of using either region alone (ablation) | GAIC cropping module · GAICD test | average SRCC: 0.706 -> 0.735; 0.597 -> 0.735 | Helped |
| Use a sufficiently large aligned spatial resolution; very small resolutions (3x3 or 5x5) lower performance (ablation) | GAIC cropping module · GAICD test | average SRCC: 0.689; 0.711; 0.725; 0.735; 0.736 | Helped |
| Reduce channel dimension from 512 to 8 rather than retaining a large feature dimension (ablation) | GAIC cropping module · GAICD test | average SRCC: 0.733; 0.736; 0.735 | No clear effect |
| Use a feature extraction layer with a larger receptive field, but avoid excessive downsampling (ablation) | GAIC feature extractor · GAICD test | average SRCC: 0.695; 0.715; 0.735; 0.737; 0.702 | Mixed |
| Use VGG16 instead of ResNet50 as the feature extractor (ablation) | GAIC feature extractor · GAICD test | average SRCC: 0.735; 0.737 vs. 0.709; 0.712 | Helped |
| Train with densely annotated candidate crops rather than pairwise ranking annotations | GAIC · GAICD test | average SRCC: 0.621 -> 0.735 | Helped |
| Use iterative A2-RL rather than the proposed GAIC approach | GAIC · GAICD test | Acc_1/5: 53.5 vs. 23.0 | Hurt |
Closest papers by meaning
- Rethinking Image Cropping: Exploring Diverse Compositions From Global Views (CVPR 2022 · similarity 0.89)
- Spatial-Semantic Collaborative Cropping for User Generated Content (AAAI 2024 · similarity 0.88)
- Cropper: Vision-Language Model for Image Cropping through In-Context Learning (CVPR 2025 · similarity 0.86)
- Learning Subject-Aware Cropping by Outpainting Professional Photos (AAAI 2024 · similarity 0.86)
- Venus: Benchmarking and Empowering Multimodal Large Language Models for Aesthetic Guidance and Cropping (CVPR 2026 · similarity 0.82)
- Object-level Geometric Structure Preserving for Natural Image Stitching (AAAI 2025 · similarity 0.81)
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.