Location Is All You Need: Continuous Spatiotemporal Neural Representations of Earth Observation Data
This provides a user-friendly alternative to Geospatial Foundation Models for end-users by eliminating data access and preprocessing overhead, though it is incremental as it builds on existing neural representation methods.
The paper tackles the problem of modeling multi-temporal Earth observation data by introducing LIANet, a coordinate-based neural representation that reconstructs satellite imagery from spatial and temporal coordinates, achieving competitive performance in downstream tasks like semantic segmentation without requiring access to original data.
In this work, we present LIANet (Location Is All You Need Network), a coordinate-based neural representation that models multi-temporal spaceborne Earth observation (EO) data for a given region of interest as a continuous spatiotemporal neural field. Given only spatial and temporal coordinates, LIANet reconstructs the corresponding satellite imagery. Once pretrained, this neural representation can be adapted to various EO downstream tasks, such as semantic segmentation or pixel-wise regression, importantly, without requiring access to the original satellite data. LIANet intends to serve as a user-friendly alternative to Geospatial Foundation Models (GFMs) by eliminating the overhead of data access and preprocessing for end-users and enabling fine-tuning solely based on labels. We demonstrate the pretraining of LIANet across target areas of varying sizes and show that fine-tuning it for downstream tasks achieves competitive performance compared to training from scratch or using established GFMs. The source code and datasets are publicly available at https://github.com/mojganmadadi/LIANet/tree/v1.0.1.