Research map / Label-efficient and robust learning / Recognition and long-tail
Embedding Label Structures for Fine-Grained Feature Representation
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 |
|---|---|---|---|
| Jointly optimize softmax classification and triplet similarity losses | GoogleNet · Stanford Car | classification accuracy: 86.9% -> 88.4% | Helped |
| Jointly optimize softmax classification and triplet similarity losses | GoogleNet · Car-333 | classification accuracy: 87.9% -> 89.4% | Helped |
| Jointly optimize softmax classification and triplet similarity losses | GoogleNet · Food dataset | classification accuracy: 87.1% -> 89.0% | Helped |
| Use triplet loss alone instead of softmax classification | GoogleNet · Stanford Car | classification accuracy: 86.9% -> 78.7% | Hurt |
| Use triplet loss alone instead of softmax classification | GoogleNet · Car-333 | classification accuracy: 87.9% -> 61.2% | Hurt |
| Use triplet loss alone instead of softmax classification | GoogleNet · Food dataset | classification accuracy: 87.1% -> 78.2% | Hurt |
| Embed hierarchical labels with generalized triplets | Joint softmax and triplet learning · Stanford Car, coarse-level retrieval | precision at top-100: at least 12.4% higher precision | Helped |
| Embed hierarchical labels with generalized triplets | Joint softmax and triplet learning · Car-333, middle-level retrieval | precision at top-2000: at least 13.2% better | Helped |
| Embed hierarchical labels with generalized triplets | Joint softmax and triplet learning · Car-333, top-level retrieval | precision at top-2000: 12.8% better; 7.2% better than ours without embedding structures | Helped |
| Embed shared attributes using adaptive triplet margins | Joint softmax and triplet learning · Food dataset, fine-grained retrieval | precision at top-50: nearly the same performance | No clear effect |
| Embed shared attributes using adaptive triplet margins | Joint softmax and triplet learning · Food dataset, attribute-level retrieval | precision at top-50: 3.1% better | Helped |
| Set the softmax loss weight to 0 or 1 rather than jointly weighting both losses (ablation) | Joint softmax and triplet learning · Fine-grained datasets | feature representation and classification | Mixed |
Closest papers by meaning
- Navigating Real-World Partial Label Learning: Unveiling Fine-Grained Images with Attributes (AAAI 2024 · similarity 0.86)
- Fine-grained Classes and How to Find Them (ICML 2024 · similarity 0.86)
- GUIDED: Granular Understanding via Identification, Detection, and Discrimination for Fine-Grained Open-Vocabulary Object Detection (NeurIPS 2025 · similarity 0.85)
- FLAIR: VLM with Fine-grained Language-informed Image Representations (CVPR 2025 · similarity 0.85)
- Democratizing Fine-grained Visual Recognition with Large Language Models (ICLR 2024 · similarity 0.84)
- Advancing Fine-Grained Classification by Structure and Subject Preserving Augmentation (NeurIPS 2024 · 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.