Research map / Vision-language and multimodal / Human and embodied modeling
Learning to Taste: A Multimodal Wine Dataset
Thoranna Bender, Simon Moe Sørensen, Alireza Kashani, Kristjan Eldjarn Hjorleifsson, Grethe Hyldig, Søren Hauberg, Serge Belongie, Frederik Rahbæk Warburg
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 |
|---|---|---|---|
| add flavor modality to image+text embeddings (ablation) | FEAST (CLIP+t-SNE+CCA) · WineSensed coarse flavor attribute prediction (avg accuracy, 5-fold CV) | accuracy (SVM): 0.22 -> 0.28 | Improved |
| add text to image+flavor (ablation) | FEAST · WineSensed coarse flavor attribute prediction (avg accuracy, 5-fold CV) | accuracy (SVM): 0.23 -> 0.28 | Improved |
| add flavor to text (ablation) | FEAST · WineSensed coarse flavor attribute prediction (avg accuracy, 5-fold CV) | accuracy (SVM): 0.16 -> 0.23 | Improved |
| add flavor to text (ablation) | FEAST · WineSensed coarse flavor attribute prediction (avg accuracy, 5-fold CV) | accuracy (NN): 0.18 -> 0.18 | No clear effect |
| use NMDS instead of t-STE for human kernel (ablation) | FEAST · WineSensed coarse flavor attribute prediction (avg accuracy, 5-fold CV) | accuracy (SVM): 0.13 -> 0.16 | Improved |
| use t-SNE instead of UMAP for machine kernel reduction (ablation) | FEAST · WineSensed coarse flavor attribute prediction (avg accuracy, 5-fold CV) | accuracy (NN): 0.18 -> 0.25 | Improved |
| use t-SNE instead of PCA for machine kernel reduction (ablation) | FEAST · WineSensed coarse flavor attribute prediction (avg accuracy, 5-fold CV) | accuracy (SVM): 0.20 -> 0.22 | Improved |
| use CCA instead of SNaCK to combine kernels (ablation) | FEAST · WineSensed coarse flavor attribute prediction (avg accuracy, 5-fold CV) | accuracy (SVM): 0.23 -> 0.28 | Improved |
| use CCA instead of ICP to combine kernels (ablation) | FEAST · WineSensed coarse flavor attribute prediction (avg accuracy, 5-fold CV) | accuracy (NN): 0.24 -> 0.26 | Improved |
| use image encoders instead of text encoders (ablation) | machine kernel · WineSensed coarse flavor attribute prediction (avg accuracy, 5-fold CV) | accuracy | Hurt |
| add flavor (human kernel NMDS+CCA) to CLIP+t-SNE text (ablation) | FEAST · WineSensed fine-grained (TAR) | TAR: 0.82 -> 0.88 | Improved |
| add flavor to CLIP+t-SNE image (ablation) | FEAST · WineSensed fine-grained (TAR) | TAR: 0.82 -> 0.89 | Improved |
Closest papers by meaning
- OVFoodSeg: Elevating Open-Vocabulary Food Image Segmentation via Image-Informed Textual Representation (CVPR 2024 · similarity 0.85)
- Recipe1M+: A Dataset for Learning Cross-Modal Embeddings for Cooking Recipes and Food Images (TPAMI 2019 · similarity 0.84)
- CHEF: Cross-modal Hierarchical Embeddings for Food Domain Retrieval (AAAI 2021 · similarity 0.84)
- Learning Cross-Modal Embeddings for Cooking Recipes and Food Images (CVPR 2017 · similarity 0.84)
- Low-Rank Similarity Mining for Multimodal Dataset Distillation (ICML 2024 · similarity 0.81)
- Pic2Prep: A Multimodal Conversational Agent for Cooking Assistance (AAAI 2025 · similarity 0.80)
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.
