Figure 3
From: Deep learning applied to hyperspectral endoscopy for online spectral classification
Classification results for the tube dataset, corresponding to a scan of a color chart area. (a) Schematic of the process of acquiring tube images. (b) Stitched version of the reference image, corrected for the fibre artifacts and stitching artifacts. (c) Prediction map of the scanned area. (d) Proposed real-time feedback for the slices (i, ii, iii) highlighted on the prediction map. A typical output is comprised of the white-light background image, with the classified hyperspectral data being shown in the line from which it was collected.
