AI Driven Age and Gender Estimation from Digital Barefoot prints: An Android Application

Abstract

The forensic value of barefoot impressions lies in the fact that they may contain anatomical and morphological characteristics which can be useful as a preliminary demographical assessment if standard identification evidence is not found or not appropriate. Typical “barefoot-print” analysis, however, typically requires a manual anthropometric measurement and judgment from the clinician, and this process can be time consuming and observer-dependent. This study proposed and validated an artificial intelligence (AI)-Based Mobile System for Age and Gender estimation of digital barefoot images for Forensic Pattern Recognitation perspective. An experimental design using two phases was used. As a first step, the development of FootAI Prediction application, an Android based application that combines a digital image acquisition component, cropping of plantar region, preprocessing of the acquired images, and Morphopod AI prediction model, running on TensorFlow Lite, was developed. To assess the application, a total of 200 orthogonal digital barefoot-print images were collected for Phase II and 1,000 barefoot-print samples were used for developing the underlying model. Images were used to standardize, isolate the plantar regions of interest, analyze the plantar images by AI based morphology, and classify the images with regard to verified demographic data. The application was able to complete the analytical workflow without failing, and successfully predicted age group in the independent validation with an accuracy of 69%. Automated classification across pre-defined age-ranges was showcased in representative cases resulting in predictions in seconds, for male/female and across age categories. The results show the potential of integrating mobile computing with an AI-based machine learning image retrieval system for initial demographic inferences of digital "barefoot" images. Yet, it will be necessary to validate these with larger and richer data, with actual crime scenes impressions. FootAI is generally a promising Forensic AI application for rapid demographic analysis that is portable.