One Computer Science MSc thesis to be presented on 4 November

Monday, 4 November there will be a master thesis presentation in Computer Science at Lund University, Faculty of Engineering.

The presentation will take place in E:4130 (LUCAS).

Note to potential opponents: Register as an opponent to the presentation of your choice by sending an email to the examiner for that presentation (firstname.lastname@cs.lth.se). Do not forget to specify the presentation you register for! Note that the number of opponents may be limited (often to two), so you might be forced to choose another presentation if you register too late. Registrations are individual, just as the oppositions are! More instructions are found on this page.

14:15-15:00 in E:4130 (LUCAS)

Presenter: Alexander Hedre
Title: Evaluation of audio event detection machine learning models for gunshots in offline browser client
Examiner: Elin A. Topp
Supervisors: Noric Couderc (LTH), Axel Nilsson (Axis Communications AB), Alex Gustafsson (Axis Communications AB)

This thesis explores the feasibility of implementing real-time gunshot detection in an offline, browser-based environment using machine learning models with the TensorFlow.js framework. The study addresses Axis Communication's need for a solution to assist in monitoring body camera audio feeds in scenarios where manual supervision is impractical and no access to a cloud based environment is available.

Two audio event detection models, AST and YAMNet, were adapted to TensorFlow.js with a neural network on their output trained for gunshot detection. Various configurations were tested to ensure real-time processing was possible without disrupting the user interface.

The results demonstrate the viability of running gunshot detection models offline with sufficient throughput while not disrupting other processes. Additionally, it highlights the potential for adapting other model types for the same environment.

Keywords: Gunshot Detection, Audio Event Detection, Machine Learning, Model Adaptation, TensorFlow.js, Offline Browser Application, Real-time Processing.

Link to popular science summary: https://fileadmin.cs.lth.se/cs/Education/Examensarbete/Popsci/241104_14Hedre.pdf