-
1
Colorectal cancer has a 47% fatality rate, with early diagnosis crucial for improving survival, yet only 40% are detected early.
-
2
Monocular depth estimation is vital for endoscopic 3D reconstruction, aiding in polyp detection and navigation despite visibility challenges.
-
3
The proposed SHADeS model improves depth estimation by jointly estimating depth, albedo, shading, and specular reflections, outperforming existing methods.
-
4
SHADeS effectively decouples albedo from specular reflections, reducing artefacts that hinder accurate depth estimation in endoscopic images.
-
5
The model can also generate specularity segmentation masks and inpaint images, enhancing the quality of endoscopic visualizations.
