Wei Ji Leong
Geospatial Data Scientist
Research Interests
- Mapping the distribution of subglacial water in Antarctica
- Deep Learning on large Earth Observation datasets
- Reproducible Science via Open Source Software
Geospatial Data Science Skills
Deep Learning & Remote Sensing:
- Computer Vision: Comfortable with designing Convolutional Neural Networks for image segmentation of satellite imagery. Very knowledgeable about state-of-the-art models including Generative Adversarial Networks for super resolution, and have dabbled with actor-critic Reinforcement Learning models. Well acquainted with Bayesian-based hyperparameter tuning and modifying neural network architectures for better performance on novel applications.
- Satellite Remote Sensing: Very proficient with processing ICESat/GLAS & ICESat-2/ATLAS laser altimeter point cloud data for measuring surface elevation changes. Familiar with using optical (e.g. Landsat 5/7/8 and Sentinel-2) and SAR (e.g. Sentinel-1) imagery for monitoring land cover changes. Also adept at LiDAR and aerial imagery processing workflows.
- Reproducible Research: Practical experience with structuring machine learning projects from dataset preprocessing to groundtruth validation. Adept at using dvc to version control training data and log experimental artifacts. Very capable of producing visually pleasing data visualizations in static and interactive formats for a range of audiences.
Programming/Scripting:
- Python: Excellent proficiency, from running vectorized NumPy operations, to handling 2D Pandas tables and n-dimensional Xarray datasets. Able to handle large datasets via Dask for distributed CPU computing and CuPy/Rapids AI for GPU accelerated workflows. Adept at utilizing various geospatial Python libraries like PyGMT, Geopandas and Rasterio. Avid user of Pytorch/Chainer/Keras/Tensorflow for prototyping deep neural networks.
- Cloud computing: Informed about Cloud-native Geospatial concepts such as Cloud-Optimized GeoTIFFs and the STAC specification. Experienced with using Microsoft Planetary Computer (Azure), NVIDIA NGC and Google Cloud Platform to prototype applications. Capable of running asynchronous processing scripts using SLURM and kbatch.
Open source mentality:
- Active contributor on public Github projects like zen3geo, PyGMT and xbatcher. Likes to keep code maintainable by writing quality unit tests and automating repeatable processes using continuous integration/deployment services.
- High competence in using Jupyter notebooks/QGIS as a platform for analyzing spatial data; strong familiarity (> 10 years experience) with Linux command-line interface.
- Familiar with using git to track code and document changes; docker containers and mamba/conda/pip to ensure reproducible environments
Tertiary Education
- Antarctic Research Centre, Victoria University of Wellington, New Zealand
- PhD in Glaciology/Physical Geography (2017-2021).
- BSc (Hons), First Class Honours in Geology, with selection of other papers in Remote Sensing, Physical Geography and Environmental Psychology (2015).
- BSc, majoring in Environmental Science and Geography, including some notable Geological and Biological science components (2012-2014).
Community Experience
-
~ Presentation ~ Cloud-Native Geospatial Webinar - Pacific Perspectives
-
~ Presentation ~ The ecosystem of geospatial machine learning tools in the Pangeo world
-
~ Workshop ~ zen3geo: Guiding Earth Observation data on its path to enlightenment
-
~ Workshop ~ EGU22 Short Course: Crafting beautiful maps with PyGMT
-
to ~ ICESat-2 Hackweek - Data Visualization tutorial - Making nice maps for posters with Python
-
to ~ ESWN Professional Development Workshop: Developing Free and Open Source Software with PyGMT
-
~ Workshop ~ PyGMT for geoscientists: A PyData compatible package for analyzing and plotting time-series and gridded data
Academic Experience
-
Poster - Applying geometry-aware Clifford Fourier Neural Operators to weather forecasting Leong, W. J., & Roy, S. (2024, April 9). Applying geometry-aware Clifford Fourier Neural Operators to weather forecasting. MIGARS, Wellington, New Zealand. https://doi.org/10.6084/m9.figshare.25599978
-
Presentation - DeepSlide: Self-supervised learning on SAR data for change detection Leong, W. J., Mahesh, R. B., Prapas, I., Boehm, V., Ramos-Pollan, R., Nemni, E., Kalaitzis, F., & Ganju, S. (2022, December 16). DeepSlide: Self-supervised learning on SAR data for change detection. AGU 2022 Fall Meeting, Chicago, IL & Online Everywhere.
-
Poster - H2OvalNet: Detecting Fairy Circles from Sentinel-2 imagery in Australia Leong, W. J., Yadav, B., Negrete, A., Howat, I., & Moortgat, J. (2022, December 13). H2OvalNet: Detecting Fairy Circles from Sentinel-2 imagery in Australia. AGU 2022 Fall Meeting, Chicago, IL & Online Everywhere. https://doi.org/10.22541/essoar.167214400.07921060/v1
-
Presentation - DeepBedMap: A Super-Resolution Generative Adversarial Network for resolving the subglacial topography of Antarctica Leong, W. J., & Horgan, H. J. (2022, June 17). DeepBedMap: A Super-Resolution Generative Adversarial Network for resolving the subglacial topography of Antarctica. Machine Learning for Polar Regions Workshop, Lamont-Doherty Earth Observatory, virtual. https://www.climate.columbia.edu/events/machine-learning-polar-regions-workshop
-
Presentation - Teaching machines about our planet: Viewing, Learning, Imagining Leong, W. J. (2022, April 08). Teaching machines about our planet: Viewing, Learning, Imagining. OSU EARTHSC 8898 Seminar, virtual. https://earthsciences.osu.edu/events/wei-ji-leong-earthsc-8898-teaching-machines-about-our-planet-viewing-learning-imagining
-
Poster - Collaborative Computational Resource Development around ICESat-2 Data: the icepyx Community and Library Scheick, J., Bisson, K., Li, T., Leong, W. J., & Arendt, A. A. (2021, December 17). Collaborative Computational Resource Development around ICESat-2 Data: the icepyx Community and Library. AGU 2021 Fall Meeting, virtual. https://doi.org/10.1002/essoar.10511316.1
-
E-lightning - PyGMT: An open-source Python library for geospatial processing, analysis, and visualization Jones, M., Grund, M., Schlitzer, W., Leong, W. J., Tian, D., Yao, J. & Uieda, L. (2021, December 17). PyGMT: An open-source Python library for geospatial processing, analysis, and visualization. AGU 2021 Fall Meeting, virtual.
-
Poster - Spatiotemporal variability of active subglacial lakes in Antarctica from 2018-2021 using ICESat-2 laser altimetry Leong, W. J., & Horgan, H. J. (2021, December 14). Spatiotemporal variability of active subglacial lakes in Antarctica from 2018-2021 using ICESat-2 laser altimetry. AGU 2021 Fall Meeting, virtual. https://doi.org/10.1002/essoar.10510996.1
-
Thesis - The subglacial landscape and hydrology of Antarctica mapped from space Leong, W. J. (2021). The subglacial landscape and hydrology of Antarctica mapped from space [PhD thesis, Victoria University of Wellington | Te Herenga Waka]. https://doi.org/10.26686/wgtn.14956062.v1
-
Poster - Spatiotemporal variability of active subglacial lakes in Antarctica from 2018-2020 using ICESat-2 laser altimetry Leong, W. J., & Horgan, H. J. (2021, February 10). Spatiotemporal variability of active subglacial lakes in Antarctica from 2018-2020 using ICESat-2 laser altimetry. New Zealand Antarctic Science Conference, Christchurch, New Zealand. https://doi.org/10.13140/RG.2.2.27952.07680
-
Paper - DeepBedMap: A deep neural network for resolving the bed topography of Antarctica Leong, W. J., & Horgan, H. J. (2020). DeepBedMap: A deep neural network for resolving the bed topography of Antarctica. The Cryosphere, 14(11), 3687–3705. https://doi.org/10.5194/tc-14-3687-2020
-
Presentation - DeepBedMap: A Super-Resolution Generative Adversarial Network for resolving the bed topography of Antarctica Leong, W. J., & Horgan, H. J. (2020, October 20). DeepBedMap: A Super-Resolution Generative Adversarial Network for resolving the bed topography of Antarctica. CEDSG Seminar Series, virtual. https://talks.cam.ac.uk/talk/index/152497
-
Poster - DeepBedMap: Resolving the bed topography of Antarctica with a deep neural network Leong, W. J., & Horgan, H. J. (2020, October 6). DeepBedMap: Resolving the bed topography of Antarctica with a deep neural network. SCAR 2020, virtual. https://doi.org/10.1002/essoar.10506291.1
-
Presentation - DeepBedMap: A super-resolution deep neural network for resolving the bed topography of Antarctica Leong, W. J., & Horgan, H. J. (2019, July 10). DeepBedMap: A super-resolution deep neural network for resolving the bed topography of Antarctica. IGS Five Decades of Radioglaciology Symposium, Stanford, California, United States. https://www.igsoc.org/symposia/2019/stanford/proceedings/procsfiles/procabstracts_75.html#A3032
-
Poster - DeepBedMap: Using a deep neural network to better resolve the bed topography of Antarctica Leong, W. J., & Horgan, H. J. (2019, April 9). DeepBedMap: Using a deep neural network to better resolve the bed topography of Antarctica. EGU General Assembly, Vienna, Austria. https://presentations.copernicus.org/EGU2019/EGU2019-11945_presentation.pdf