MIT-Harvard CINCS (Communications Information Networks Circuits and Signals) / Hamilton Institute Seminar
Speaker: Professor Rozenn Dahyot, Trinity College Dublin
Title: "AI Design, deployment & applications for Digital Twins & Sustainability"
Abstract: Data driven software engineering using deep learning has been a very active area of research for the past decade or so.
This talk will present recent works done in my team for using convolutional neural networks (CNNs) as part of
intelligent systems for addressing climate change events (e.g. floods), sustainability (e.g. harvesting solar energy efficiently) and infrastructure deployment & management (e.g. telecom).
In addition to CNNs, these pipelines also include graphs capturing explicit (explainable) prior rules for information fusion and decision, and current efforts aim at improving CNNs computational efficiency, their collaboration with graph structures as well as exploring alternative design with robust and learnable activation functions.
IM2ELEVATION: Building height estimation from single-view aerial imagery
Remote Sensing (2020) DOI:10.3390/rs12172719
Automatic detection of passable roads after floods in remote sensed and social media data
Signal Processing: Image Communication (2019) DOI:10.1016/j.image.2019.02.002
Bonseyes AI Pipeline - bringing AI to you. End-to-end integration of data, algorithms and deployment tools
ACM Transactions on Internet of Things (2020) DOI:10.1145/3403572
Automatic Discovery and Geotagging of Objects from Street View Imagery
Remote Sensing (2018) DOI:10.3390/rs10050661
Harmonic Convolutional Networks based on Discrete Cosine Transform
Semantic image segmentation based on spatial relationships and inexact graph matching
International Conference on Image Processing Theory, Tools and Applications (IPTA 2020),
L2 Divergence for Robust Colour Transfer
Computer Vision and Image Understanding (2019)