Taylor Denouden Email & Phone Number

Currently working at the Hakai Institute located in Victoria, BC where my role is to produce semantic segmentation models for marine species detection. Our goal is to reduce the human effort required to label and delineate species like kelp and seagrass in UAV and aerial imagery. This work involves creating suitable training datasets, implementing and optimizing deep learning models, and setting up training and deployment pipelines using local and cloud-based accelerated computing hardware.My graduate research at the University of Waterloo was focussed on safety in autonomous vehicle perception systems. I worked with the Autonomoose team supervised by Dr. Krzysztof Czarnecki and investigated methods for detecting unusual objects in object detection perception systems. My research showed that some reduction of false-positive detection was possible in two-stage object detectors, but that it was difficult to reduce the number of false-negative detections. Other projects at University of Waterloo included a system for improving out-of-distribution detection using the latent feature-space of autoencoder networks, and research comparing the efficacy of various anomaly detection systems for fraudulent credit card transactions.Previously, I worked at the Hakai Research Institute in the IT and GIS department in Victoria, British Columbia, where I helped to support scientific research staff with their data storage and visualization needs. My work included the design, implementation, and maintenance of databases, creation of interactive visualizations for exploratory data analysis, and automation of quality assurance and processing workflows. Before Hakai, I worked as a developer on the BikeMaps.org project with Dr. Trisalyn Nelson at the University of Victoria's Spatial Analysis and Research Laboratory in Victoria, BC.

Listed skills include Gis, Programming, Leaflet, D3.Js, and 16 others.