benjaminfraser Goto Github PK

I'm Ben, a Aerospace Engineer (Avionics, Electronic and Electrical Systems) and Data Science Practitioner.

I have a specialist focus on Applied AI techniques, Machine Learning and Data Science Technologies. My experience in software, data science and programming extends over a decade, with expertise developed on a range of professional and personal projects in that time. I love working with technology, learning new skills, and working across a variety of programming languages. Currently my main efforts focus around Python-based development.

Development of an Integrated Crowd-Monitoring and Social Distancing Platform that exploits video surveillance data and novel comuter-vision techniques. Application of Deep Computer Vision, including Crowd-Pose Estimation, Object Detection, Target Tracking and Automated Homography. Research published through MFI2022 conference. Source code available: https://github.com/BenjaminFraser/Social-Distancing-Pose-Platform.

Development of a Digital-Twin based Novelty Detection Framework for Unmanned Aerial Systems, through the application of Deep Unsupervised Learning (Autoencoder-based) techniques. Published through IEEE as part of DASC2021 conference: https://ieeexplore.ieee.org/document/9594321/.

A NLP application that develops a Multi-Modal Text (NLP) and Tabular Classification Framework to help analysts streamline the process of performing reviews on incoming applications. Developed as part of an MVP product to enhance the productivity within a key MoD governmental department. Application of classical and deep ML techniques in Tensorflow, with a focus on Natural Language Processing.

Development of an Object Detection framework for the detection and classification of objects within satellite imagery, with a focus on Defence context. This included an end-to-end framework developed in PyTorch, with custom fine-tuned Faster-RCNN and YOLOv5 variants to work with satellite based imagery. Key features included a bespoke preprocessing phase that better prepared satellite imagery objects for classification through tiling and data-augmentation.

Development of an Object Detection Model for the analysis of Face Mask usage within complex scenes. Built using a Faster RCNN object detector fine-tuned in PyTorch using a custom mask object detection dataset, with a focus on complex scenes that are representative of real world applications and camera surveillance feeds. Once the model was trained and optimised for the task, an inference dashboard was developed in Dash to allow new predictions to be made on selected images. Source code and project summary available at: https://github.com/BenjaminFraser/Mask-Object-Detection.

An Image Classification Web Application, built using Flask, Bootstrap and JS to allow for the uploading of a given image that is subsequently classified. The preparation and formation of a deep convolutional network model was performed, making use of transfer-learning and fine-tuning of the Inception architecture. The application will classify images according to the specific transfer learning task carried out (the app was demonstrated with different dog breeds). Source code available at: https://github.com/BenjaminFraser/Dog-Breed-Classifier-App.

A Python GUI application that uses Tkinter, matplotlib, numpy, scipy and other scientific computing libraries to provide classical and digital electronic and control engineering functions for graphical analysis. Developed whilst conducting a BSc in Electronics Systems Engineering, to support cohorts (and wider electronic engineering community via GitHub) in forming control engineering plots and visualisations. Available at: https://github.com/BenjaminFraser/Control_engineering_apps.

A Python Application implemented in Flask with LibreOffice backend to convert a large (or small) directory of word and excel documents into a single PDF document. The final PDF is ordered according to the names of the directories in the uploaded zip file, and each directory in-turn is ordered numberically by the files within. User access is required on the application before any finals can be amended or converted. Available at: https://github.com/BenjaminFraser/word-to-pdf

An Intrusion Monitoring System built using a combination of many technologies, both hardware and software, to create a robust and reliable system that makes detections of intruders or motion at a chosen location. For the system, a combination of Arduinos and Raspberry Pi Devices are used to develop a network of communication nodes (using nRF24L01+ radio transceivers) and radar detection nodes (using custom-built HB100 X-Band Radar Doppler motion sensors).

The system continuously monitors the regions where the nodes are deployed, and sends an alarm to the central system on detection of an intrusion. All detections and the status of each region are monitored using a Flask web application, which provides real-time status updates on detections. Source code available at: https://github.com/BenjaminFraser/Intrusion_monitoring_system.

Blog post available at: https://medium.com/@benjamindavidfraser/arduino-nrf24l01-communications-947e1acb33fb

Centre for Autonomous and Cyber-Physical Systems, Cranfield University, United Kingdom.

Topics included Statistical Learning, Deep Learning, Computer Vision, Search & Optimisation, Intelligent Agents, Bayesian Techniques, Systems Engineering, Coding and Development (Python & MATLAB), Data Visualisation and Analytics, Reinforcement Learning, AI Research. Publication of multiple formal papers in the field of Applied-AI through the IEEE. Presentation of research at two international technology conferences.

Open University, Milton Keynes, United Kingdom.

Final dissertation on Stress Mechanics and Design included multi-dimensional Finite Element Analysis (FEA), with application of scientific and numerical computing in Python.

University of Portsmouth, United Kingdom

Two major projects undertaken, including an electronic system design, and an aviation system research project based on Electronic Warfare and Infrared Missile technologies.

📫 You can find me on Twitter or LinkedIn.