FRIDAY, AUGUST 7, 2026
FRIDAY, AUGUST 7, 2026
16:30–18:00 IST • 13:00–14:30 CEST • 07:00–08:30 EDT
16:30–18:00 IST • 13:00–14:30 CEST • 07:00–08:30 EDT
16:30–18:00 IST • 13:00–14:30 CEST • 07:00–08:30 EDT
The model session will review the technical approach used to estimate child anthropometric indicators from standard smartphone images. The AIMM team will present the computer vision pipeline, current validation results, target outcomes, and planned development of height, MUAC, weight, and nutrition-risk estimates. Participants will critically assess model validity, robustness, uncertainty, fairness, and suitability for deployment on low-specification smartphones. The discussion will focus on the evidence needed to demonstrate reliable performance across age groups, field conditions, devices, and population subgroups.
The model session will review the technical approach used to estimate child anthropometric indicators from standard smartphone images. The AIMM team will present the computer vision pipeline, current validation results, target outcomes, and planned development of height, MUAC, weight, and nutrition-risk estimates. Participants will critically assess model validity, robustness, uncertainty, fairness, and suitability for deployment on low-specification smartphones. The discussion will focus on the evidence needed to demonstrate reliable performance across age groups, field conditions, devices, and population subgroups.
Register
Register
SESSION MEMO
SESSION MEMO
Join webinar
Join webinar
Tell us your questions and suggestions:
Key Questions
Key Questions
Which deep-learning approaches could most effectively move AIMM beyond conventional regression and improve the estimation of anthropometric measurements from smartphone images?
Which deep-learning approaches could most effectively move AIMM beyond conventional regression and improve the estimation of anthropometric measurements from smartphone images?
AIMM addresses the need for faster, lower-cost, and more accurate child nutrition screening, particularly in hard-to-reach, crisis-affected, or otherwise resource-constrained settings. Current nutrition assessments often require trained personnel, physical measurement tools, and repeated field visits, which can limit coverage during emergencies, displacement, or pandemic-related restrictions. AIMM explores whether smartphone-based images can help estimate key indicators of child undernutrition and support low-cost, near-real-time identification of children at risk.
What additional data, annotations, model architectures, or training strategies would be needed to develop a robust end-to-end AI pipeline?
What additional data, annotations, model architectures, or training strategies would be needed to develop a robust end-to-end AI pipeline?
AIMM is designed for settings where reliable nutrition data are urgently needed but difficult to collect. This includes conflict-affected areas, remote communities, humanitarian emergencies, public health crises, and other situations where access to children and households is limited. By enabling smartphone-based, community-level monitoring, AIMM can help generate timely information when standard measurement systems are disrupted, overstretched, or unavailable.
Which fairness metrics and subgroup sample sizes are necessary to detect clinically meaningful performance disparities?
Which fairness metrics and subgroup sample sizes are necessary to detect clinically meaningful performance disparities?
The app is being developed so that caregivers and families can monitor child nutrition using a standard smartphone, with minimal equipment and guidance. Drawing on prior experience with emoji-based questionnaires and culturally appropriate interface designs in Kenya, AIMM will place strong emphasis on intuitive, low-literacy, and locally appropriate user interaction. The app interface will be refined together with rural households and caregivers to ensure that it is easy to understand and use in everyday settings.
Are the project’s current indicators of success appropriate, and how should clinical, operational, equity, and cost-effectiveness outcomes be prioritized?
Are the project’s current indicators of success appropriate, and how should clinical, operational, equity, and cost-effectiveness outcomes be prioritized?
AIMM is being developed and tested in Maharashtra, India. In the first phase, the study will sample approximately 7,000 children across five districts — Dhule, Chandrapur, Nagpur, Jalgaon, and Beed — to train and validate the app. In a second phase, AIMM will be tested in Pune district with 150 households over a period of 12 months.
FRIDAY, AUGUST 7, 2026
FRIDAY, AUGUST 7, 2026
16:30–18:00 IST • 13:00–14:30 CEST • 07:00–08:30 EDT
16:30–18:00 IST • 13:00–14:30 CEST • 07:00–08:30 EDT
16:30–18:00 IST • 13:00–14:30 CEST • 07:00–08:30 EDT
The model session will review the technical approach used to estimate child anthropometric indicators from standard smartphone images. The AIMM team will present the computer vision pipeline, current validation results, target outcomes, and planned development of height, MUAC, weight, and nutrition-risk estimates. Participants will critically assess model validity, robustness, uncertainty, fairness, and suitability for deployment on low-specification smartphones. The discussion will focus on the evidence needed to demonstrate reliable performance across age groups, field conditions, devices, and population subgroups.
The model session will review the technical approach used to estimate child anthropometric indicators from standard smartphone images. The AIMM team will present the computer vision pipeline, current validation results, target outcomes, and planned development of height, MUAC, weight, and nutrition-risk estimates. Participants will critically assess model validity, robustness, uncertainty, fairness, and suitability for deployment on low-specification smartphones. The discussion will focus on the evidence needed to demonstrate reliable performance across age groups, field conditions, devices, and population subgroups.
Register
Register
SESSION MEMO
SESSION MEMO
Join webinar
Join webinar
Tell us your questions and suggestions:
Key Questions
Key Questions
Which deep-learning approaches could most effectively move AIMM beyond conventional regression and improve the estimation of anthropometric measurements from smartphone images?
Which deep-learning approaches could most effectively move AIMM beyond conventional regression and improve the estimation of anthropometric measurements from smartphone images?
AIMM addresses the need for faster, lower-cost, and more accurate child nutrition screening, particularly in hard-to-reach, crisis-affected, or otherwise resource-constrained settings. Current nutrition assessments often require trained personnel, physical measurement tools, and repeated field visits, which can limit coverage during emergencies, displacement, or pandemic-related restrictions. AIMM explores whether smartphone-based images can help estimate key indicators of child undernutrition and support low-cost, near-real-time identification of children at risk.
What additional data, annotations, model architectures, or training strategies would be needed to develop a robust end-to-end AI pipeline?
What additional data, annotations, model architectures, or training strategies would be needed to develop a robust end-to-end AI pipeline?
AIMM is designed for settings where reliable nutrition data are urgently needed but difficult to collect. This includes conflict-affected areas, remote communities, humanitarian emergencies, public health crises, and other situations where access to children and households is limited. By enabling smartphone-based, community-level monitoring, AIMM can help generate timely information when standard measurement systems are disrupted, overstretched, or unavailable.
Which fairness metrics and subgroup sample sizes are necessary to detect clinically meaningful performance disparities?
Which fairness metrics and subgroup sample sizes are necessary to detect clinically meaningful performance disparities?
The app is being developed so that caregivers and families can monitor child nutrition using a standard smartphone, with minimal equipment and guidance. Drawing on prior experience with emoji-based questionnaires and culturally appropriate interface designs in Kenya, AIMM will place strong emphasis on intuitive, low-literacy, and locally appropriate user interaction. The app interface will be refined together with rural households and caregivers to ensure that it is easy to understand and use in everyday settings.
Are the project’s current indicators of success appropriate, and how should clinical, operational, equity, and cost-effectiveness outcomes be prioritized?
Are the project’s current indicators of success appropriate, and how should clinical, operational, equity, and cost-effectiveness outcomes be prioritized?
AIMM is being developed and tested in Maharashtra, India. In the first phase, the study will sample approximately 7,000 children across five districts — Dhule, Chandrapur, Nagpur, Jalgaon, and Beed — to train and validate the app. In a second phase, AIMM will be tested in Pune district with 150 households over a period of 12 months.
FRIDAY, AUGUST 7, 2026
FRIDAY, AUGUST 7, 2026
16:30–18:00 IST • 13:00–14:30 CEST • 07:00–08:30 EDT
16:30–18:00 IST • 13:00–14:30 CEST • 07:00–08:30 EDT
16:30–18:00 IST • 13:00–14:30 CEST • 07:00–08:30 EDT
The model session will review the technical approach used to estimate child anthropometric indicators from standard smartphone images. The AIMM team will present the computer vision pipeline, current validation results, target outcomes, and planned development of height, MUAC, weight, and nutrition-risk estimates. Participants will critically assess model validity, robustness, uncertainty, fairness, and suitability for deployment on low-specification smartphones. The discussion will focus on the evidence needed to demonstrate reliable performance across age groups, field conditions, devices, and population subgroups.
The model session will review the technical approach used to estimate child anthropometric indicators from standard smartphone images. The AIMM team will present the computer vision pipeline, current validation results, target outcomes, and planned development of height, MUAC, weight, and nutrition-risk estimates. Participants will critically assess model validity, robustness, uncertainty, fairness, and suitability for deployment on low-specification smartphones. The discussion will focus on the evidence needed to demonstrate reliable performance across age groups, field conditions, devices, and population subgroups.
Register
Register
SESSION MEMO
SESSION MEMO
Join webinar
Join webinar
Tell us your questions and suggestions:
Key Questions
Key Questions
Which deep-learning approaches could most effectively move AIMM beyond conventional regression and improve the estimation of anthropometric measurements from smartphone images?
Which deep-learning approaches could most effectively move AIMM beyond conventional regression and improve the estimation of anthropometric measurements from smartphone images?
AIMM addresses the need for faster, lower-cost, and more accurate child nutrition screening, particularly in hard-to-reach, crisis-affected, or otherwise resource-constrained settings. Current nutrition assessments often require trained personnel, physical measurement tools, and repeated field visits, which can limit coverage during emergencies, displacement, or pandemic-related restrictions. AIMM explores whether smartphone-based images can help estimate key indicators of child undernutrition and support low-cost, near-real-time identification of children at risk.
What additional data, annotations, model architectures, or training strategies would be needed to develop a robust end-to-end AI pipeline?
What additional data, annotations, model architectures, or training strategies would be needed to develop a robust end-to-end AI pipeline?
AIMM is designed for settings where reliable nutrition data are urgently needed but difficult to collect. This includes conflict-affected areas, remote communities, humanitarian emergencies, public health crises, and other situations where access to children and households is limited. By enabling smartphone-based, community-level monitoring, AIMM can help generate timely information when standard measurement systems are disrupted, overstretched, or unavailable.
Which fairness metrics and subgroup sample sizes are necessary to detect clinically meaningful performance disparities?
Which fairness metrics and subgroup sample sizes are necessary to detect clinically meaningful performance disparities?
The app is being developed so that caregivers and families can monitor child nutrition using a standard smartphone, with minimal equipment and guidance. Drawing on prior experience with emoji-based questionnaires and culturally appropriate interface designs in Kenya, AIMM will place strong emphasis on intuitive, low-literacy, and locally appropriate user interaction. The app interface will be refined together with rural households and caregivers to ensure that it is easy to understand and use in everyday settings.
Are the project’s current indicators of success appropriate, and how should clinical, operational, equity, and cost-effectiveness outcomes be prioritized?
Are the project’s current indicators of success appropriate, and how should clinical, operational, equity, and cost-effectiveness outcomes be prioritized?
AIMM is being developed and tested in Maharashtra, India. In the first phase, the study will sample approximately 7,000 children across five districts — Dhule, Chandrapur, Nagpur, Jalgaon, and Beed — to train and validate the app. In a second phase, AIMM will be tested in Pune district with 150 households over a period of 12 months.