Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Zero-Shot Ingredient Recognition by Multi-Relational Graph Convolutional Network
Jingjing Chen, Liangming Pan, Zhipeng Wei, Xiang Wang, Chong-Wah Ngo, Tat-Seng Chua
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 |
|---|---|---|---|
| combine hierarchy, attribute and co-occurrence graphs instead of hierarchy only (ablation) | mRGCN · VIREO Food-172 zero-shot, seen+unseen | Hit@1: 8.0 -> 23.5 | Improved |
| add co-occurrence relation to hierarchy/attribute graphs (ablation) | mRGCN · VIREO Food-172 zero-shot, seen+unseen | Hit@10: 23.4 -> 43.6 | Improved |
| add attribute relation to hierarchy+co-occurrence (ablation) | mRGCN · VIREO Food-172 zero-shot, seen+unseen | Hit@1: 21.9 -> 23.5 | Improved |
| add hierarchy relation to attribute+co-occurrence (ablation) | mRGCN · VIREO Food-172 zero-shot, seen+unseen | Hit@1: 21.5 -> 23.5 | Improved |
| combine hierarchy and attribute relations only (no co-occurrence) (ablation) | mRGCN · VIREO Food-172 zero-shot, seen+unseen | Hit@1: 8.0 -> 3.6 (vs hierarchy only) | Hurt |
| concatenation instead of sum to fuse relations (ablation) | mRGCN · VIREO Food-172 and UEC Food-100 | Hit@k | Improved |
| attention instead of concatenation to fuse relations (ablation) | mRGCN · UEC Food-100 | Hit@10: more than 12% improvement | Improved |
| attention instead of concatenation to fuse relations (ablation) | mRGCN · VIREO Food-172 | Hit@10: around 2% | Improved |
| use ResNet-50 instead of VGG-16 backbone (ablation) | mRGCN · VIREO Food-172 zero-shot, unseen only | Hit@10: 43.2 -> 47.4 | Improved |
| use averaged BERT embeddings instead of Word2Vec (ablation) | mRGCN · VIREO Food-172 | Hit@k | Hurt |
| use mRGCN instead of GCNZ | zero-shot ingredient recognition · VIREO Food-172 zero-shot, seen+unseen | Hit@10: 23.1 -> 43.6 | Improved |
| use mRGCN instead of Fast0Tag | zero-shot ingredient recognition · VIREO Food-172 zero-shot, unseen only | Hit@20: 67.7 vs 60.6 | Hurt |
Closest papers by meaning
- Fine-Grained Image Classification by Exploring Bipartite-Graph Labels (CVPR 2016 · similarity 0.86)
- Large Scale Visual Food Recognition (TPAMI 2023 · similarity 0.85)
- Learn More for Food Recognition via Progressive Self-Distillation (AAAI 2023 · similarity 0.85)
- Multi-Label Image Recognition With Graph Convolutional Networks (CVPR 2019 · similarity 0.84)
- GKGNet: Group K-Nearest Neighbor based Graph Convolutional Network for Multi-Label Image Recognition (ECCV 2024 · similarity 0.84)
- OVFoodSeg: Elevating Open-Vocabulary Food Image Segmentation via Image-Informed Textual Representation (CVPR 2024 · similarity 0.83)
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.
