Segmenting across places: The need for fair transfer learning with satellite imagery

Miao Zhang, Harvineet Singh, Lazarus Chok, Rumi Chunara

Abstract: The increasing availability of high-resolution satellite imagery has enabled the use of machine learning to support land-cover measurement and inform policy-making. How- ever, labelling satellite images is expensive and is available for only some locations. This prompts the use of transfer learning to adapt models from data-rich locations to others. Given the potential for high-impact applications of satel- lite imagery across geographies, a systematic assessment of transfer learning implications is warranted. In this work, we consider the task of land-cover segmentation and study the fairness implications of transferring models across lo- cations. We leverage a large satellite image segmentation benchmark with 5987 images from 18 districts (9 urban and 9 rural). Via fairness metrics we quantify disparities in model performance along two axes – across urban-rural locations and across land-cover classes. Findings show that state-of-the-art models have better overall accuracy in ru- ral areas compared to urban areas, through unsupervised domain adaptation methods transfer learning better to ur- ban versus rural areas and enlarge fairness gaps. In analy- sis of reasons for these findings, we show that raw satellite images are overall more dissimilar between source and tar- get districts for rural than for urban locations. This work highlights the need to conduct fairness analysis for satel- lite imagery segmentation models and motivates the devel- opment of methods for fair transfer learning in order not to introduce disparities between places, particularly urban and rural locations.