# Poshan Lens

- **Event:** [Meta Global AI Developer Hackathon](https://cerebralvalley.ai/e/meta-ai-developer-hackathon)
- **When:** Oct 26, 12:00 PM – Nov 4, 9:00 PM (EST)
- **Where:** Online
- **Team:** [Sankeerth Nara](https://cerebralvalley.ai/u/Sankeerth_Nara)
- **GitHub:** https://github.com/SankeerthNara/meta-hackathon-2026
- **Demo video:** https://www.youtube.com/watch?v=wlhVzGrMies
- **Gallery:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2740

Poshan Lens measures a young child's height with an ordinary Android phone, then tells an Anganwadi worker and the parent what the measurement means. India has about 1.4 million Anganwadi centres, and their workers are meant to measure every child under five every month. Height is the number that reveals stunting, which affects 35.5 % of Indian children under five (NFHS-5), and it is the measurement most often skipped or done wrong.

How it works: the worker types the child's age (and weight from the scale), points the phone at the standing child, and slides it slowly left and right for nine seconds. Chrome's WebXR (ARCore) records the camera frames and the exact phone motion. SAM 3.1 finds the child in every frame. Pluto, my motion-measurement method, turns the masks and the phone motion into a height in centimetres. It takes three independent measurements from different parts of the sweep and keeps the value at least two agree on. Code then computes the WHO Child Growth Standards z-scores and categories (severe / moderate / normal). Muse Spark checks one frame (is the whole body visible, is the child standing straight?) and writes a summary and checklist for the worker, a message for the parent in their own language and script (Hindi, Telugu, Tamil, ...), read aloud on the phone, and a note for the growth register. It needs no LiDAR, no depth camera, no measuring board and no app install, and costs about $0.15 of API per child. It is a screening aid that tells the worker whom to refer, not a diagnosis.

## More from Meta Global AI Developer Hackathon

- [Verified Asset](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2737)
- [RiskLens](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2738)
- [Bull & Moose Party](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2739)
- [MuseVibe](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2741)
- [Future Builders](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2742)
- [Show Me Once: Verified Receiving Agent](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2743)

---

Markdown version of https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2740. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt