Overview

NutriScreener is an AI-assisted malnutrition screening framework developed jointly by IIT Jodhpur and AIIMS Jodhpur, published at the AAAI Conference on Artificial Intelligence (AAAI-26, AI for Social Impact Track).

The system integrates retrieval-augmented learning with multi-pose graph attention networks (GATs) to assess malnutrition risk and predict anthropometric measures (height, weight, MUAC, head circumference) from natural child images.
It is designed for scalable, low-cost screening in community and clinical environments, where manual measurements may be infeasible.

The underlying knowledge base (MalKB) and toolkit adhere to Responsible AI principles, emphasizing privacy, interpretability, and clinician supervision.
All data were collected under institutional ethics approvals, and the released versions use non-reversible CLIP embeddings, ensuring no identifiable visual content is stored or shared.

Released Components (Research Use Only)

Under the NutriScreener research license, the following components may be provided to eligible institutions:

• Evaluation scripts and configuration files
• Model weights for NutriScreener (Weighted)
• MalKB embeddings and/or Data
  A curated pediatric knowledge base derived from consented participants
• (Optional) NutriScreener research toolkit
  The developed toolkit UI for research and educational purposes only.
  
  

All components are provided strictly for non-commercial, research and educational use. Detailed clauses and responsibilities are specified in the license agreement.

Ethical and Privacy Note.

• Any extension of the knowledge base (MalKB) in your own studies should follow your local ethics / IEC approvals.
  
  
• NutriScreener is a screening research tool, not a stand-alone diagnostic system, and it’s being released Open Source for research purposes only

To obtain the password for the compressed file(s) containing the NutriScreener resources, email the duly filled license agreement to  databases@iab-rubric.org with the subject line "Signed license agreement for NutriScreener and MalKB".

NOTE:
The license agreement has to be signed by someone having the legal authority to sign on behalf of the institute, such as the head of the institution or registrar. If a license agreement is signed by someone else, it will not be processed further.

These resources are available only for research and educational purposes and not for any commercial use.

If you use NutriScreener, MalKB, or any of the released components in your publications or reports, you must refer to the following paper:

M. Khan, M. Vatsa, R. Singh, K. Singh.
NutriScreener: Retrieval-Augmented Multi-Pose Graph Attention Network for Malnourishment Screening.
Proceedings of the AAAI Conference on Artificial Intelligence (AI for Social Impact Track), 2026.