Research map / Label-efficient and robust learning / Recognition and long-tail
Retrieving and Classifying Affective Images via Deep Metric Learning
Read the paperCheck 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 |
|---|---|---|---|
| replace FC feature with sentiment vector embedding (ablation) | softmax + sentiment loss · FI | accuracy (%): 65.84 -> 67.64 | Helped |
| replace FC feature with sentiment vector embedding (ablation) | softmax + sentiment loss · FI | mAP8: 0.4575 -> 0.4885 | Helped |
| replace triplet constraint with sentiment constraint (ablation) | softmax + metric loss, semi-hard sampling · FI | mAP8: 0.4426 -> 0.4575 | Helped |
| replace triplet constraint with sentiment constraint (ablation) | softmax + metric loss, semi-hard sampling · FI | mAP2: 0.7592 -> 0.7913 | Helped |
| add softmax loss to triplet loss (ablation) | GoogleNet triplet · FI | accuracy (%): 63.46 -> 65.35 | Helped |
| semi-hard sampling instead of random sampling (ablation) | softmax + triplet · FI | mAP8: 0.4016 -> 0.4426 | Helped |
| semi-hard sampling instead of hard sampling (ablation) | softmax + triplet · FI | mAP8: 0.4228 -> 0.4426 | Helped |
| use metric learning loss alone instead of softmax (ablation) | GoogleNet triplet · FI | accuracy (%): accuracy 65.18 -> 63.46, mAP8 0.3583 -> 0.3951 | Mixed |
| vary loss weight omega within 0.6-0.8 (ablation) | multi-task loss · FI | accuracy (%): within 0.5% difference | No clear effect |
| deeper CNN architecture (AlexNet to VGG to GoogleNet) | softmax fine-tuning · FI | accuracy (%): 58.13 -> 64.55 -> 65.18 | Helped |
Closest papers by meaning
- MDAN: Multi-Level Dependent Attention Network for Visual Emotion Analysis (CVPR 2022 · similarity 0.88)
- Building a Large Scale Dataset for Image Emotion Recognition: The Fine Print and The Benchmark (AAAI 2016 · similarity 0.87)
- Affection: Learning Affective Explanations for Real-World Visual Data (CVPR 2023 · similarity 0.87)
- AffectGPT: A New Dataset, Model, and Benchmark for Emotion Understanding with Multimodal Large Language Models (ICML 2025 · similarity 0.86)
- MART: Masked Affective RepresenTation Learning via Masked Temporal Distribution Distillation (CVPR 2024 · similarity 0.86)
- FindingEmo: An Image Dataset for Emotion Recognition in the Wild (NeurIPS 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.