Charles Explorer logo
🇬🇧

Towards Assembling Photo Lineup Identification via Convolutional Neural Networks

Publication at Faculty of Mathematics and Physics, Faculty of Arts |
2017

Abstract

Lineups may be used as key evidence in the investigation process, especially in cases where eyewitness testimony is the only evidence. This project focuses specifically on creating lineups based on photographs - a photo lineup.

The main goals of the project are to create practical tools, data sets, and guidelines that will contribute to a simpler, more accurate administration of photo lineups and their subsequent verifications. We intend to achieve this goal with machine learning methods, which are based on learning the similarity of different objects through their pictures.

Outputs of the machine learning methods will be compared with photo lineups created by experts in terms of lineup fairness.