Research map / Graph learning / Knowledge and temporal graphs
NutriScreener: Retrieval Augmented Multi-Pose Graph Attention Network for Malnourishment Screening
Misaal Khan, Mayank Vatsa, Kuldeep Singh, Richa Singh
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 retrieval augmentation with class-boosted weighting and context fusion to CLIP+GNN (ablation) | NutriScreener · AnthroVision | malnutrition recall: 0.54 -> 0.79 | Improved |
| add retrieval augmentation to CLIP+GNN (ablation) | NutriScreener · AnthroVision | height RMSE (cm): 7.37 -> 6.38 | Improved |
| use retrieval alone without the GAT (ablation) | NutriScreener · AnthroVision | AUC: 0.82 -> 0.61 | Hurt |
| fine-tune the CLIP encoder instead of keeping it frozen (ablation) | NutriScreener · AnthroVision | malnutrition recall: 79% -> 38% | Hurt |
| fine-tune the CLIP encoder instead of keeping it frozen (ablation) | NutriScreener · AnthroVision | ROC-AUC: 0.82 -> 0.72 | Hurt |
| replace weighted BCE with focal loss (ablation) | NutriScreener (BCE) · AnthroVision | F1: 0.59 -> 0.53 | Hurt |
| add class-aware temperature-scaled retrieval weighting on top of context fusion (ablation) | NutriScreener (Context) · AnthroVision | AUC: 0.78 -> 0.82 | Improved |
| add class-aware temperature-scaled retrieval weighting on top of context fusion (ablation) | NutriScreener (Context) · AnthroVision | height RMSE (cm): 11.10 -> 6.38 | Improved |
| add context signals (log-odds, local density) to fusion with BCE (ablation) | NutriScreener (BCE) · AnthroVision | precision: 0.47 -> 0.54 | Improved |
| use a demographically matched KB (ARAN-H2) for ARAN-H1 (ablation) | NutriScreener · ARAN-H1 (regression only) | Ht RMSE (cm): 7.87 -> 7.21 | Improved |
| use MalKB instead of no retrieval (ablation) | NutriScreener · AnthroVision test | recall: 0.54 -> 0.79 | Improved |
| use CampusPose (out-of-distribution) KB for retrieval (ablation) | NutriScreener · CampusPose | Ht RMSE (cm): 23.98 -> 23.98 | No clear effect |
Closest papers by meaning
- NutriBench: A Dataset for Evaluating Large Language Models in Nutrition Estimation from Meal Descriptions (ICLR 2025 · similarity 0.83)
- EatVid-Bench: A Multimodal Fine-Grained Eating Behavior Video Dataset (ECCV 2026 · similarity 0.83)
- OmniFood8K: Single-Image Nutrition Estimation via Hierarchical Frequency-Aligned Fusion (CVPR 2026 · similarity 0.83)
- MedTrinity-25M: A Large-scale Multimodal Dataset with Multigranular Annotations for Medicine (ICLR 2025 · similarity 0.80)
- Medical thinking with multiple images (ICLR 2026 · similarity 0.80)
- CLIMB: Data Foundations for Large Scale Multimodal Clinical Foundation Models (ICML 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.
