Fig. 1: Illustration of the proposed model.
From: A human-machine collaborative approach measures economic development using satellite imagery
The model is composed of machine-driven clustering of satellite images (Stage 1), human guidance on the partial order graph (POG) of image clusters (Stage 2), and a machine-driven rank-wise score model to compute siScore (Stage 3). A POG contains information on the relative ranking of each cluster’s development, perceived and judged by each human expert. Knowledge from multiple POGs is summarized as a single representative POG using an ensemble process. The result is then used to train the score model at the subsequent stage. Satellite images in the figure contain modified Copernicus Sentinel data [2016, 2017, 2018, 2019].
