Here is some of the formal research I have done or been part of academically (not in industry):
MiDaS: A large-scale Minecraft dataset for non-natural image benchmarking
The Journal of Electronic Imaging paper is available here.
On the Robustness of Self-Supervised Representations for Multi-view Object Classification
The Pattern Recognition Letters paper is available here.
SummaryNet: A Multi-Stage Deep Learning Model for Automatic Video Summarisation
The arXiv pre-print is available here.
Human Action Recognition using Local Two-Stream Convolutional Neural Network Features and Support Vector Machines
The arXiv pre-print is available here.
Masters of Science
My MSc. Computer Science research involved finding a novel algorithm to human action recognition. This included desiging a pipeline involving novel preprocessing techniques, 3D CNNs, optical flow correction, and linear SVMs. Earlier in the degree, I conducted research into steganography; machine learning on homomorphically-encrypted data; and image compression using autoencoders and JPEG2000.
Image Compression using Autoencoders
The goal of this research project was to perform compression of hyperspectral imagery using autoencoders and JPEG2000 (an improved variant of the JPEG image compression standard). The joint latent space of the autoencoder (i.e. for all channels of the image) was fed into JPEG2000 for compression. The compression was more effective since the AE could efficiently decorrelate the channels. The JPEG2000 bit stream was used as the compressed representation.
Automated Parking Detection
This research was my Honours year project in university. I gathered the data manually. I performed an investigation of multiple potential real-time techniques for automated parking detection. These included neural networks, Viola-Jones cascade, and a simple RGB-based multivariate Gaussian approach.