John Quinn

John Quinn

John Quinn is a researcher with the Google AI team in Ghana. His interests are in the application of machine intelligence methods to global challenges, e.g. in health, agriculture, and disaster response, and for addressing information gaps in the developing world. He is also a faculty member in the Department of Computer Science at Makerere University in Uganda, where he has worked since 2007. For five years prior to joining Google, he was the technical lead on a number of Africa analytics projects for United Nations Global Pulse, a UN data science initiative. He received a BA in Computer Science from the University of Cambridge in 2000, and PhD in machine learning from the University of Edinburgh in 2007.

Research Areas

Authored Publications
Sort By
  • Title
  • Title, descending
  • Year
  • Year, descending
    High Resolution Building and Road Segmentation from Sentinel-2 Imagery
    Abdoulaye Diack
    Abel Tesfaye Korme
    Emmanuel Asiedu Brempong
    Jason Hickey
    Juliana Marcos
    Krishna Sapkota
    Mohammed Alewi Hassen
    Wojciech Sirko
    arXiv, https://arxiv.org/abs/2310.11622 (2023)
    Preview abstract Mapping buildings and roads automatically with remote sensing typically requires imagery of at least 50 cm resolution, which is expensive to obtain and often sparsely available. In this work we demonstrate how public, worldwide imagery from the Sentinel-2 Earth observation mission can be used to carry out this task at a much higher level of detail than the 10 m raw pixel resolution would suggest. To do this, we employ a teacher-student method in which a model with access to a temporal stack of Sentinel-2 images is trained to make the same predictions as a high-resolution model with access to corresponding 50 cm imagery. Evaluating at 50cm resolution, we achieve mIOU of 0.78, equivalent in accuracy to applying a single-frame high resolution model with imagery of 4m resolution. This work opens up new possibilities for using freely available Sentinel-2 imagery for a range of downstream tasks that previously could only be done with high resolution satellite imagery. The model will be made available soon to non-commercial, non-governmental entities at https://sites.research.google/open-buildings/ upon request. View details
    Preview abstract Identifying the locations and footprints of buildings is vital for many practical and scientific purposes, and such information can be particularly useful in developing regions where alternative data sources may be scarce. In this work, we describe a model training pipeline for detecting buildings across the entire continent of Africa, given 50cm satellite imagery. Starting with the U-Net model, widely used in satellite image analysis, we study variations in architecture, loss functions, regularization, pre-training, self-training and post-processing that increase instance segmentation performance. Experiments were carried out using a dataset of 100k satellite images across Africa containing 1.75M manually labelled building instances, and further datasets for pre-training and self-training. We report novel methods for improving performance of building detection with this type of model, including the use of mixup (mAP +0.12) and self-training with soft KL loss (mAP +0.06). The resulting pipeline obtains good results even on a wide variety of challenging rural and urban contexts, and was used to create the Open Buildings dataset of approximately 600M Africa-wide building footprints. View details