Key Information

Register
Submit
The challenge is finished.

Challenge Overview

Background:

Worldwide, one in four children is stunted. Three-quarters of them live in South Asia and sub-Saharan Africa.


Photographer: +rex  Photo Sourced:  https://www.flickr.com/photos/_rex/1441506201/in/gallery-97445835@N07-72157639158747604/  
Creative Commons License: 
https://creativecommons.org/licenses/by/2.0/legalcode

Stunting not only affects a child’s height. It also has an impact on brain development. Stunted children are more likely to fall behind at school, miss key milestones in reading and math, and go on to live in poverty. When stunted children don’t reach their potential, neither do their countries. Growth faltering saps a country’s strength, lowering productivity and keeping the entire nation trapped in poverty.

While the extreme hunger and starvation that once defined sub-Saharan Africa are now rare, today the issue isn’t quantity of food as much as it is quality—whether kids are getting enough protein and other nutrients to fully develop physically and mentally. The global health community is still working to understand all of the causes and solutions to physical and cognitive growth stunting and failure to thrive. The problem is complex and involves quantification of the interactions between nutrition quantity and quality, good health, sanitation and hygiene, education and strong nurturing family and social networks, etc.

The ability to make accurate measurements of physical growth is a core element in studying the effects of stunting in a population and resolving this critical issue.

Challenge Goal:

Our goal is to explore the ability to develop an Android application that enables frequent and accurate collection of length/height measures of infants, toddlers, children, adolescents and adults.  Gains in length (children under the age of 2 are measured lying down) and height (children above 2 years old are measured standing up).  You may use any of the sensors on phone including, of course, the camera.

It’s fair to assume that we can ask the user to identify the top/bottom of the child in the image, as well as potential body parts (“tap on the center of the child’s head/face”).  Accurate breakdowns of length measurements into leg length, torso, shoulder width and sitting height gain extra points.

The app will be used in several countries and in extremely rural areas, and so the traditional approach of using a reference object to calculate length appears to be a challenge. Take into account that the mobile device might be at different heights and angles when taking the photos that will be analyzed.

We would like you to submit a strategy for an algorithm or combination of algorithms that can be used based on inputs from the phone (camera, accelerometer, any of the sensors…) to determine the length/height of the child with reasonable accuracy. You may also take more than 1 photo, in combination with an app which directs/instructs the user. We want you to document very creative ideas which will enable us towards our goal, together with a little bit of hard work. You are not required to demonstrate a working solution - just tell us how to get there.

The submissions should demonstrate the feasibility of this app, along with descriptions of the algorithms and processes that would make this possible.  Although, you don’t have to provide code for this “ideation” type challenge, your solution will be judged on the feasibility of the solution and the detail provided.

Technical Background:

There are several potential approaches to this problem:

  1. Taking multiple pictures to create a three dimensional model of a subject.  The pictures might be distributed by depth or taken in slightly different positions across a plane equidistant from the subject.   The images could be processed with Fixed Model or Mixed Model effects.

  2. Using a reference object.

  3. Using trigonometry, the camera’s accelerometer and the camera’s height to estimate the height of subjects photographed.

  4. Using one of the camera’s other sensors to detect distance from a subject.

There are some applications in the Google Play store which use some of these principles:

The problem with these tools is their lack of accuracy.  Testing of the tools has shown that height of objects in the pictures is only accurate to +/-5%.  Other articles to spur your creativity:

There is also a zip file attached with various scholarly articles about the problem domain.  These are not endorsements of a particular solution.   Rather, they are background information so that your research isn’t starting from ground-zero.



Final Submission Guidelines

This challenge is being launched as a code challenge because we want input from software developers but actually there isn’t any code required.  You’ll be submitting a document outlining your proposed solution to this issue.  You’ll be graded (subjectively) by the feasibility of your solution and the level of detail you offer.  You may include supplemental material that helps explain your idea such as screenshots or mockups. You are encouraged to provide references to literature or publications where needed or appropriate.

Target Audience

- International Health Workers

Judging Criteria

In order of importance, you will be judged on:

  • How feasible is your idea?  

  • Level of detail provided?

ELIGIBLE EVENTS:

2015 topcoder Open

REVIEW STYLE:

Final Review:

Community Review Board

Approval:

User Sign-Off

SHARE:

ID: 30049199