We leverage dynamic graph neural networks (DGNN) for computer vision (CV) tasks to study remote sensing (RS) imagery, including denoising, classification, or prediction. Moreover, DGNN also provide more explainable decomposition to CV models. Depending on temporal resolution, autoregression (AR), recurrent neural network (RNN) or long short-term memory (LSTM) may be used. Available datasets (e.g., agricultural RS) are open source.

Prerequisites:
Basics of Statistics, GNN
PyTorch