Brandon Birmingham

B.Sc.(Hons)(Melit.), M.Sc.(Melit.)

Research Developer

About

myphoto
  • Name: Brandon Birmingham
  • Nationality: Maltese
  • Age:

A highly motivated, determined and hardworking IT enthusiast with proven abilities to meet agreed deadlines and competence in any assigned work. Always focused on any task at hand and able to utilise my theoretical knowledge and practical experience to be successful as possible in any challenging problem. I am constantly on the lookout to improve and enhance my knowledge with new technologies and cutting-edge research in Machine Learning and Computer Vision.

Research Interests

  • Machine Learning
  • Deep Learning
  • Computer Vision
  • Digital Forensics

Areas of Expertise

  • Web Development
  • Software Engineering
  • Data Science
  • Business Intelligence
signature

Resume

Education

  • 2017 - present

    Doctoral degree

    Ph.D. ICT, University of Malta

    Investigating how machine learnt models for sub-tasks in automatic generation of image captions can enhance the generated or retrieved image descriptions. This research specifically aims to study how object detection, content selection, generation of prepositions, verb prediction and scene recognition can be exploited for retrieving and synthesising textual descriptions extracted from textual data found on the Web.

  • 2015 - 2016

    Postgraduate degree

    M.Sc. Computer Science, University of Malta

    Dissertation: "Predicting Spatial Relationships for Image Descriptions".

  • 2012 - 2015

    Graduate Degree

    B.Sc. (Hons.) in Computing Science, University of Malta

    Final Year Project: "Smart Carving: Recovering Deleted JPEG Files". This project was graded A, awarded a top industry award and recognised as the best ICT Project 2015 by the Chamber of Engineers.

  • 2010 - 2012

    Matriculation Certificate

    G.F. Abela Junior College, Msida

    Advanced Level: Pure Maths (A), Computing (A)
    Intermediate Level: Physics (A), Religion (A), Maltese (B), SOK (B)
    Overall grade: A

  • 2005 - 2010

    School Leaving Certificate

    Junior Lyceum, Hamrun

    Awarded for best student behavior, student of the month and for academic achievements.

Experience

  • Jul 2017 - Present

    Senior BI Developer

    Eunoia

    Responsible for the development, implementation and support of Business Intelligence solutions based on Microsoft technologies as well as for the overall engineering and implementation of a web-based financial consolidation application. Technologies: Microsoft SQL Server, Microsoft SSIS, SSRS, SSAS, C#, ASP.NET MVC, Python.

  • Dec 2016 - Jul 2017

    BI Developer

    PTL Ltd.

    Responsible for the implementation of a web-based financial consolidation tool which supports a BI infrastructure for reporting and visualisation. Technologies: Microsoft SQL Server, Microsoft SSIS, SSRS, SSAS, C#, ASP.NET MVC, Python.

  • Oct 2016 - Nov 2016

    Conducting research in the area of automated image captioning which involves the integration of Computer Vision and Natural Language Processing. Specifically, this project involved the development of a deep learning model which combines language and automatically generated geometric features to predict spatial prepositions. Technologies: Python, Keras

  • Oct 2015 - Jan 2016

    Preparing and delivering C programming hands-on tutorial sessions for first year ICT University students. The tutorial sessions were intended to help students troubleshoot their code implemented in C as preparation for their study unit's assigned project.

  • Jun 2012 - Aug 2016

    Software Developer

    HOB Software Malta

    Team Lead and Full Stack Web developer responsible for an internal web application that serves as a project management framework. I was also introduced to one of the main products of the company, a secure VPN solution, in which I was responsible for the re-design and re-implementation of the IKE protocol of both client's and gateway's solution. Technologies used: Java, JSP, DB2 SQL, HTML5, CSS3, JavaScript, jQuery, Ajax, C and C++.

Skills

Software Development

95%

Web Development

80%

Data Science

80%

Business Intelligence

70%

Personal Skills

Focus
Hardworking
Dedication
Analytical
Communication

4

Years of Experience

6

Publications

8

Certificates

7

Awards

Publications

image-file-carving

B. Birmingham, R. A. Farrugia, M. Vella. Using Thumbnail Affinity for Fragmentation Point Detection of JPEG Files. In IEEE Proceedings EUROCON, pages 3-8, Ohrid, Macedonia, July 6-8, 2017. (Best Student Paper Award)

Abstract: File carving tools carry out file recovery whenever the file-system meta-data is not available, which makes them a valuable addition to the cyber crime investigator’s toolkit. Existing file carvers either cannot handle fragmented files or require a probabilistic model derived using a number of training images. This training data may not always be feasible to aggregate or its sheer size could undermine practicality. Similar to existing techniques, our method exploits both the JPEG syntax and semantic-based analysis steps in order to distinguish the correct fragments required for recovering images. The thumbnail affinity-based semantic analysis constitutes the novel aspect of this approach. Comparative evaluation using three widely used benchmark test sets show that our carver compares with the state-of-the-art commercial tool that requires an a-priori model while beating a number of popular forensic tools. This outcome demonstrates the successful replacement of the probabilistic model with thumbnail affinity, rendering this technique the right complement for existing carvers in situations where thumbnail information is readily available.

image-captioning

B. Birmingham, A. Muscat. The Use of Object Labels and Spatial Prepositions as Keywords in a Web-Retrieval-Based Image Caption Generation System. In Proceedings of the 6th Workshop on Vision and Language. Association for Computational Linguistics, pages 11-20, Valencia, Spain, April 4, 2017.

Abstract: In this paper, a retrieval-based caption generation system that searches the web for suitable image descriptions is studied. Google’s search-by-image is used to find potentially relevant web multimedia content for query images. Sentences are extracted from web pages and the likelihood of the descriptions is computed to select one sentence from the retrieved text documents. The search mechanism is modified to replace the caption generated by Google with a caption composed of labels and spatial prepositions as part of the query’s text alongside the image. The object labels are obtained using an offthe-shelf R-CNN and a machine learning model is developed to predict the prepositions. The effect on the caption generation system performance when using the generated text is investigated. Both human evaluations and automatic metrics are used to evaluate the retrieved descriptions. Results show that the web-retrieval-based approach performed better when describing single-object images with sentences extracted from stock photography websites. On the other hand, images with two image objects were better described with template-generated sentences composed of object labels and prepositions.

image-file-carving

B. Birmingham. A. Muscat. Feature generation and selection for predicting spatial relationships in 2D images. In Computer Science Annual Workshop 2016, Faculty of ICT University of Malta, November 10-11, 2016.

Abstract: The prediction of spatial relationships between objects in a two-dimensional image is a subtask in automatic image description generation that has been largely ignored until late. Research efforts so far have been in using machine learned models whose feature space consists of both language and visual hand-engineered features. This paper studies (i) the status quo in spatial preposition prediction, (ii) optimises and improves the machine learning model, and (iii) investigates how features can be automatically learned in a deep learning setup.

image-file-carving

A. Belz, A. Muscat, B. Birmingham, J. Levacher, J. Pain, A. Quinquenel. Effect of data annotation, feature selection and model choice on spatial description generation in french. In Proceedings of the 9th International Natural Language Generation. Association for Computational Linguistics, pages 237-241, Edinburgh, UK, September 5-8, 2016.

Abstract: In this paper, we look at automatic generation of spatial descriptions in French, more particularly, selecting a spatial preposition for a pair of objects in an image. Our focus is on the effect on accuracy of (i) increasing data set size, (ii) removing synonyms from the set of prepositions used for annotation, (iii) optimising feature sets, and (iv) training on best prepositions only vs. training on all acceptable prepositions. We describe a new data set where each object pair in each image is annotated with the best and all acceptable prepositions that describe the spatial relationship between the two objects. We report results for three new methods for this task, and find that the best, 75% Accuracy, is 25 points higher than our previous best result for this task.

image-file-carving

A. Muscat, A. Belz, B. Birmingham. Exploring different preposition sets, models and feature sets in automatic generation of spatial image descriptions. In Proceedings of the 5th Workshop on Vision and Language. Association for Computational Linguistics, pages 65-69, Berlin, Germany, August 12, 2016.

Abstract: In this paper we look at the question of how to create good automatic methods for generating descriptions of spatial relationships between objects in images. In particular, we investigate the impact of varying different aspects of automatic method development, including using different preposition sets, models and feature sets. We find that optimising the preposition set improves previous best Accuracy from 46.2 to 50.2. Feature set optimisation further improves best Accuracy from 50.2 to 53.25. Naive Bayes models outperform SVMs and decision trees under all conditions tested. The utility of individual features depends on the model used, but the most useful features tend to capture a property pertaining to both objects jointly.

image-prepositions

B. Birmingham. A. Muscat. Predicting spatial relationships for image descriptions using combined vision and language features. In the 8th Workshop in Information and Communication Technology, Faculty of ICT University of Malta. July 8, 2016.

Abstract: In this paper we look at the question of how to create good automatic methods for generating descriptions of spatial relationships between objects in images. In particular, we investigate the impact of varying different aspects of automatic method development, including using different preposition sets, models and feature sets. We find that optimising the preposition set improves previous best Accuracy from 46.2 to 50.2. Feature set optimisation further improves best Accuracy from 50.2 to 53.25. Naive Bayes models outperform SVMs and decision trees under all conditions tested. The utility of individual features depends on the model used, but the most useful features tend to capture a property pertaining to both objects jointly.

Get in touch

Send me a message
Address Communications & Computer Engineering
Faculty of Information & Communication Technology
Level 0, Block A
Room 6
Faculty of ICT
University of Malta
Msida