Sexing white 2D footprints using convolutional neural networks

Footprints are left, or obtained, in a variety of scenarios from crime scenes to anthropological investigations. Determining the sex footprint can be useful screening such impressions and attempts have been made do so using single multi landmark distances, shape analyses via density friction ridges. Here we explore relative importance different components sexing two-dimensional foot namely, size, texture. We use machine learning approach compare this more traditional methods discrimination. Two datasets used, pilot data set collected students at Bournemouth University (N = 196) larger by podiatrists Sheffield NHS Teaching Hospital 2677). Our convolutional neural network with accuracy around 90% on test N 267 images all image components, which is better than an expert achieve. However, quality impacts success rate, but results promising time it may possible create automated algorithm practitioners whatever sort (medical forensic) obtain first order footprint.