Research map / Vision-language and multimodal / Human and embodied modeling
Aesthetic Critiques Generation for Photos
Kuang-Yu Chang, Kung-Hung Lu, Chu-Song Chen
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 |
|---|---|---|---|
| fuse hidden annotations of aspect-specific LSTMs with soft attention (AF) instead of whole-dataset CNN-LSTM | CNN-LSTM-WD · PCCD | SPICE F-score: 0.136 -> 0.150 | Improved |
| fuse aspect-specific hidden annotations (AF) instead of selecting a single aspect model (AO) | AO Approach · PCCD | SPICE F-score: 0.127 -> 0.150 | Improved |
| train on the whole dataset (CNN-LSTM-WD) instead of one aspect only (AO) | AO Approach · PCCD | SPICE F-score: 0.127 -> 0.136 | Improved |
| train on the whole dataset (CNN-LSTM-WD) instead of one aspect only (AO) | AO Approach · PCCD | caption diversity | Hurt |
| use AF instead of AO | AO Approach · PCCD (AMT users) | Good rating share: 19.7% -> 28.8% | Improved |
| use AF instead of AO | AO Approach · PCCD (experts) | Good rating share: 10.8% -> 16.8% | Improved |
| use AF instead of AO | AO Approach · PCCD (AMT users) | Bad rating share: 8.2% -> 14.0% | Hurt |
| use AF instead of CNN-LSTM-WD | CNN-LSTM-WD · PCCD (AMT users) | average score: 2.04 -> 2.15 | Improved |
| use AF instead of CNN-LSTM-WD | CNN-LSTM-WD · PCCD (experts) | average score: 1.89 -> 2.03 | Improved |
| use AF instead of AO in comparison to ground-truth critiques | AO Approach · PCCD (vs ground truth) | generated judged better share: 51.4% -> 58.4% | Improved |
| use AF instead of AO for novel caption generation | AO Approach · PCCD | novel captions share: 37% -> 66% | Improved |
| pretrain CNN-LSTM on MSCOCO before fine-tuning on PCCD (ablation) | CNN-LSTM · PCCD | captioning performance | Improved |
Closest papers by meaning
- Understanding Aesthetics with Language: A Photo Critique Dataset for Aesthetic Assessment (NeurIPS 2022 · similarity 0.90)
- Aesthetically Relevant Image Captioning (AAAI 2023 · similarity 0.89)
- Bridging Cognitive Gap: Hierarchical Description Learning for Artistic Image Aesthetics Assessment (AAAI 2026 · similarity 0.85)
- Advancing Comprehensive Aesthetic Insight with Multi-Scale Text-Guided Self-Supervised Learning (AAAI 2025 · similarity 0.85)
- ArtiMuse: Fine-Grained Image Aesthetics Assessment with Joint Scoring and Expert-Level Understanding (CVPR 2026 · similarity 0.85)
- The Photographer's Eye: Teaching Multimodal Large Language Models to See, and Critique Like Photographers (CVPR 2025 · similarity 0.84)
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.
