Research map / Architectures and efficiency / CNN design
Effective Aesthetics Prediction With Multi-Level Spatially Pooled Features
Vlad Hosu, Bastian Goldlucke, Dietmar Saupe
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 |
|---|---|---|---|
| Use Pool-3FC with wide MLSP features instead of Single-3FC with narrow features (ablation) | Inception-v3 features on AVA · AVA test set, original-resolution images | SRCC: 0.729 -> 0.745 | Improved |
| Use Pool-3FC with wide MLSP features instead of Single-3FC with narrow features (ablation) | InceptionResNet-v2 features on AVA · AVA test set, original-resolution images | SRCC: 0.743 -> 0.756 | Improved |
| Add crop and flip augmentation (ablation) | Single-3FC on AVA rescaled images · AVA test set | SRCC: 0.668 -> 0.676 | Improved |
| Add crop and flip augmentation (ablation) | Pool-3FC with InceptionResNet-v2 features · AVA test set, original-resolution images | SRCC: 0.752 -> 0.756 | Improved |
| Use a 3FC head instead of a single fully connected layer (ablation) | Inception-v3 narrow features on AVA · AVA test set, rescaled images | SRCC: 0.644 -> 0.676 | Improved |
| Use all 11 feature blocks instead of only the last 5 (ablation) | Single-1FC on Inception-v3 features · AVA test set, rescaled images | SRCC: 0.570 -> 0.644 | Improved |
| Use separate 3FC heads for each feature block rather than one shared head (ablation) | Inception-v3 narrow features on AVA · AVA test set, rescaled images | SRCC: 0.676 -> 0.673 | No clear effect |
| Increase the Single-3FC dropout rates to (0.5, 0.5, 0.75) (ablation) | Fine-tuned Inception-v3 head on AVA · AVA test set, rescaled images | SRCC: 0.658 -> 0.675 | Improved |
| Train on original-resolution images rather than rescaled images | Proposed Pool-3FC method · AVA test set | SRCC: 0.676 -> 0.756 | Improved |
| Use pre-trained feature learning instead of end-to-end fine-tuning | Single-3FC on Inception-v3 · AVA test set, rescaled images | SRCC: 0.658 -> 0.676 | Improved |
| Use InceptionResNet-v2 rather than Inception-v3 for feature extraction | Pool-3FC on AVA original-resolution images · AVA test set | SRCC: 0.745 -> 0.756 | Improved |
| Optimize and evaluate for SRCC rather than binary accuracy | Aesthetics assessment on AVA · AVA test set | SRCC: 0.612 -> 0.756 | Improved |
Closest papers by meaning
- Image Aesthetic Assessment Assisted by Attributes through Adversarial Learning (AAAI 2019 · similarity 0.91)
- Composition-Preserving Deep Photo Aesthetics Assessment (CVPR 2016 · similarity 0.91)
- Bridging Cognitive Gap: Hierarchical Description Learning for Artistic Image Aesthetics Assessment (AAAI 2026 · similarity 0.87)
- Advancing Comprehensive Aesthetic Insight with Multi-Scale Text-Guided Self-Supervised Learning (AAAI 2025 · similarity 0.86)
- ELTA: An Enhancer against Long-Tail for Aesthetics-oriented Models (ICML 2024 · similarity 0.86)
- Unlocking the Essence of Beauty: Advanced Aesthetic Reasoning with Relative-Absolute Policy Optimization (ICLR 2026 · 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.
