Project DetailsThe BMI index is a crucial index which gives a quantitative assessment of whether a person is in normal weight, underweight or overweight. The index is calculated using height and weight data. The purpose of our project is to estimate a person's BMI from a single 2-dimensional image. This is a complex task because visual inspection of the image is not sensitive to the distance of the object from the camera and the angle of the shot. To approach this task, we relied on previous works in the field, on their results, and on the dataset published with them. In the first stage, the solution relied on a deep learning architecture designed to solve a regression problem so that we received at the output a scalar that represents BMI. Also, in order to improve the generalization capabilities of the model, we performed various augmentations on the images and tested segmentation of human body to differ it from the background. In the second stage we addressed the problem in two levels: first one is classification into 5 or more different classes according to a rough division of the BMI range, and the second is to perform a smaller regression problem in each class that will lead to BMI with greater accuracy. Another goal of the project is to run the application through a webcam and get prediction in real-time. We achieved the best results for a simple regression problem using the "efficientNet B2" neural network, in addition with different augmentations on the dataset.
