Active Learning for Hyperspectral Image Classification: A comparative review

Machine learning algorithms have demonstrated impressive results for land cover mapping from hyperspectral data. To enhance generalization capabilities of statistical models, active (AL) methods guide the annotation training data set by querying most informative samples. The classifier can then be performed on an optimal set. We bring under same framework uncertainty, representativeness, and performance-based AL techniques; conduct a benchmark state-of-the-art release toolbox ( https://github.com/Romain3Ch216/AL4EO ) to allow experimentation with these approaches. experiments are conducted various sets: toy set, classic sets, complex scene. evaluate usual accuracy metrics as well complementary metrics, which us provide guidelines when choosing relevant strategy in real use case.