My research focuses on developing deep learning tools to monitor urban areas in response to rapid urbanization's impacts on biodiversity, ecosystems, water, and public health. These tools, using Convolutional Neural Networks, Reinforcement Learning, and Generative Adversarial Networks, aid in addressing challenges and improving decision-making. The project employs Landsat (1985-2022), Sentinel-2 (2016-2022), and Planet Lab (2018-2022) images to segment urban areas using multi-spectral sensor data. The satellite-based urban classification products are crucial for managing global urbanization growth. (Examples: Landsat in resource watch, Sentinel2 in GEE, and Planet in GEE)

I am developing a machine learning system for remote sensing that combines deep learning and decision trees. The decision tree model serves as the foundation for the deep learning architecture, resulting in improved interpretation, precision, performance, and credibility. This hybrid approach allows for better regularization by discarding irrelevant metric configurations and revealing the hierarchical and non-linear relationships between variables. The goal is to create deep learning models that are both interpretable and accurate.