A total of 1800 estimates from non-expert dietitians were collected online. This comprised of 120 subjects each providing an estimate per image (120 subjects * 15 images). The overall aim of this study was to use crowdsourcing to determine the accuracy of non-experts in estimating calories in a meal as shown in a photograph. We analyzed percentiles and used bootstrapping with replacement to determine the typical percentile rank estimate that is in proximity to the ground truth and to determine the least number of sample estimates from a crowd of non-experts that can provide a convenient and accurate estimate of the number of calories in a meal.
|Title of host publication||2018 IEEE International Conference on Bioinformatics and Biomedicine(BIBM 2018)|
|Place of Publication||Madrid, Spain|
|Publication status||Published - 3 Dec 2018|