Fig. 6: One-shot hyperspectral segmentation and object tracking.

a Configuration of the general architecture of Fig. 2 for one-shot video semantic segmentation. b, c Segmentation results on hyperspectral object tracking with (b) Hyperspectral signature selected from the datacube acquired by the UAV of Fig. 4. c Segmentation results in real-time with detailed zoomed regions. d Video recording from inside a car and initially selected signatures. e A segment of hyperspectral data with signatures that the architecture tracks and propagates over time.