Multi-modal, Multi-temporal and Multi-scale SPatially-Aware Domain Adaptation framework
This repository provides code for M3SPADA framework which allows temporal domain adaptation (temporal transfer) of classification model designed to produce land cover maps based on Satellite Image Time Series Very High Resolution mono-date satellite images. The code uses Python and TensorFlow 2.X with Keras as the main model building API.
This framework is detailed in the research paper "Multi-Modal Temporal Unsupervised Domain Adaptation for Land Cover Mapping with spatial pseudo labelling and adversarial learning" which is currently under submission.
The explanations regarding the format of SITS and VHR data handled by M3SPADA are included in Python script "main_M3SPADA.py".
This project is given under the terms of the MIT license.