JASMINE

2026-05-16 · Source: Machine Learning on Medium · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Medical Devices & Health Technology, Internet of Things (IoT) & Connected Devices · Depth: Intermediate, medium

Summary

JASMINE is a low-cost, internet-free camera designed for early skin cancer detection in rural US areas, where 46 million Americans lack adequate access to physicians. The device aims to provide a pre-screening solution, leveraging a Raspberry Pi 3B+ with an 8MP Pi Camera, NeoPixel ring for lighting, and a magnifier attachment, costing approximately $145. Its software utilizes MobileNetV2 for object detection and U-Net for semantic segmentation, trained on the Skin Cancer MNIST: HAM10000 dataset. The system runs TensorFlow Lite for efficiency and uses BlueZ for Bluetooth image transfer, achieving 88% accuracy for top-2 predictions. User research, including interviews with dermatologists and a diary study with proxy users, informed design iterations, leading to a capacitive touchscreen and an Electron-based GUI.

Key takeaway

For AI hardware engineers developing medical diagnostic tools for underserved populations, JASMINE demonstrates a viable approach. Your focus should be on integrating efficient deep learning models like MobileNetV2 and U-Net onto low-cost edge devices like Raspberry Pi, ensuring offline functionality and user-friendly interfaces. Prioritize early and continuous user research to validate design choices and improve usability, especially for critical applications like cancer screening.

Key insights

Early skin cancer detection in rural areas can be significantly improved with low-cost, offline, edge AI devices.

Principles

• Prioritize user research before high-fidelity prototyping.
• Iterative design is crucial for product refinement.
• Adapt to new information and pivot project focus as needed.

Method

The JASMINE device uses MobileNetV2 for object detection and U-Net for semantic segmentation on dermatoscopic images, deployed on a Raspberry Pi 3B+ with TensorFlow Lite for offline, real-time inference and Bluetooth transfer.

In practice

• Use MobileNetV2 for efficient edge device deployment.
• Employ U-Net for biomedical image segmentation.
• Integrate Bluetooth for offline data transfer.

Topics

• Skin Cancer Detection
• Rural Healthcare Access
• Edge AI Devices
• MobileNetV2
• U-Net Segmentation

Best for: Machine Learning Engineer, Computer Vision Engineer, AI Hardware Engineer

Related on AIssential

• Spotlight on innovation: Google-sponsored Data Science for Health Ideathon across Africa — Google's Ideathon fostered African innovation in healthcare AI using open models like MedGemma and MedSigLIP.
• PRET is a few-shot system for pan-cancer recognition without example training — PRET enables flexible, scalable, and effective pan-cancer recognition with minimal examples, outperforming traditional AI and human pathologists.
• Bridging the Rural Healthcare Gap: A Cascaded Edge-Cloud Architecture for Automated Retinal Screening — An edge-cloud cascade reduces cloud usage for DR screening while maintaining high diagnostic performance.
• AI tool developed at UVM could improve Parkinson's diagnosis - Burlington Free Press — An AI tool from UVM improves Parkinson's diagnosis accuracy and speed by integrating clinical data and diagnostic criteria.
• End-user validation of BRIGHT with custom-developed graphical user interface applied to cervical cancer brachytherapy — BRIGHT with its new GUI offers an excellent, user-friendly semi-automated planning tool for cervical cancer brachytherapy.
• Scalable Training of Spatially Grounded 2D Vision-Language Models for Radiology — Spatially grounded radiology VLMs can be trained at scale without manual annotations using LLM-based curation.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Machine Learning on Medium.