Research map / Image generation / Autoregressive and latent generation
Deep Cropping via Attention Box Prediction and Aesthetics Assessment
Wenguan Wang, Jianbing Shen
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 |
|---|---|---|---|
| use the attention-box regression network instead of the best saliency model (SalNet) | attention box prediction · PASCAL attention box (850 images) | IoU: 0.379 -> 0.517 | Improved |
| use the attention-box regression network instead of a classic saliency model (ITTI) | attention box prediction · PASCAL attention box (850 images) | IoU: 0.318 -> 0.517 | Improved |
| use the proposed cascaded crop pipeline instead of the aesthetics method LCC | photo cropping · MSR-ICD (photographer 1) | IoU: 0.748 -> 0.813 | Improved |
| use the proposed cascaded crop pipeline instead of the aesthetics method LCC | photo cropping · MSR-ICD (photographer 1) | BDE: 0.066 -> 0.030 | Improved |
| use the proposed pipeline instead of sliding-window ARC aesthetics ranking | photo cropping · MSR-ICD (photographer 1) | IoU: 0.448 -> 0.813 | Improved |
| use the proposed pipeline instead of the attention-based ATC | photo cropping · MSR-ICD (photographer 1) | IoU: 0.605 -> 0.813 | Improved |
| use the proposed pipeline instead of VBC | photo cropping · FLMS | IoU: 0.74 -> 0.81 | Improved |
| use the proposed pipeline instead of ATC | photo cropping · FLMS | BDE: 0.063 -> 0.057 | Improved |
| use the proposed AA network instead of the handcrafted-feature AVA baseline | aesthetics assessment · AVA test (19,930 images) | accuracy: 0.667 -> 0.769 | Improved |
| use the proposed AA network instead of CPD | aesthetics assessment · AVA test (19,930 images) | accuracy: 0.774 -> 0.769 | No clear effect |
| use the proposed AA network instead of ARC | aesthetics assessment · AVA test (19,930 images) | accuracy: 0.773 -> 0.769 | No clear effect |
Closest papers by meaning
- A Deep Network Solution for Attention and Aesthetics Aware Photo Cropping (TPAMI 2018 · similarity 0.97)
- Image Cropping with Composition and Saliency Aware Aesthetic Score Map (AAAI 2020 · similarity 0.93)
- Spatial-Semantic Collaborative Cropping for User Generated Content (AAAI 2024 · similarity 0.86)
- Cropper: Vision-Language Model for Image Cropping through In-Context Learning (CVPR 2025 · similarity 0.86)
- Venus: Benchmarking and Empowering Multimodal Large Language Models for Aesthetic Guidance and Cropping (CVPR 2026 · similarity 0.86)
- Learning Subject-Aware Cropping by Outpainting Professional Photos (AAAI 2024 · 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.
