- Sustainable development goals
- Good health and well-being
- Zero hunger
- Project link
About the project
Child Growth Monitor (CGM) is an AI-powered smartphone app that was launched by the German non-profit organization Welthungerhilfe. The app allows individuals to scan children and detect malnutrition.
More project information
Child Growth Monitor (CGM) is an AI-powered smartphone app that was launched by the German non-profit organization Welthungerhilfe. The app allows individuals to scan children and detect malnutrition. According to its creators, the app uses an infrared sensor available in some smartphones in order to capture 3D measurements of a child’s height, body volume and weight ratio, as well as head and upper arm circumferences down to the millimeter.
Malnutrition and hunger are pervasive issues which are problematic not only because of the difficulty to provide food to different populations but also because it is a complex issue to identify. It is often the case that caretakers of children suffering from malnutrition are unable to determine that their child is suffering from it and respond too late to the problem. Additionally, the standardized measurements done by aid organisations and governmental health workers are time consuming and expensive.
Welthungerhilfe along with its partners Action Against Hunger, MGSA and Darshana Mahila Kalyan Samiti have developed the CGM in order to make the measurement process much faster and as accurate as possible. India is the first country where CGM is being piloted in three states—Madhya Pradesh, Maharashtra and Rajasthan. 150 field workers in 12 teams are supported by Action Against Hunger for using the app to scan children. Currently the team is collecting data both manually and with the app to train the artificial intelligence model to make it more accurate. Action Against Hunger’s field workers have so far scanned over 7,000 children and are aiming to complete 10,000 scans by March 2019. It is important to note that the app provides authenticated users an interface to scan children in 3D with consent of the parents and upload all collected data to the secure backend.
Because of the limitations of mobile connectivity in rural areas and in slums with tin roofs offline first is a major goal of the project. While the app already works fine in offline environments, results from the scans are currently produced in the cloud. Providing predictions directly on the device is the next big step we are taking, as it would also improve privacy by not having to upload every scan. Currently the App works through person detection and pose estimation and overlaying the information of the position of 14 points on the body of the child with the 3D point cloud from the Tango API.
The next iteration for the Pilot will guide the user to scan the child in a way that a quick, accurate measurement can be taken. This will involve data of the camera pose, point clouds and RGB video. An accurate prediction of the height of a human is priority number one. Goal is to do an 99,5% accurate prediction, so that we can measure a child of 100cm height with an error of +/- 5mm. To reconstruct a 3d model of a child or of the skeleton is a non-trivial task using multiple point clouds of a moving child. Using a single point cloud isn't accurate enough. A promising approach is to input the point clouds, the device poses for camera position and rgb video into different neural networks, to do preprocessing or get the result.
Hardware requirements for the CGM app:
Currently Google Project Tango devices only.
Will most likely expand to all devices with ARkit/ARcore capabilities (iPhone 6s and newer -100 million Android devices)
The CGM project aims to tackle malnutrition and hunger worldwide thus contributing to SDG 2 (Zero Hunger) and SDG 3 (Good Health and Well-Being).