A Benchmark for Endoluminal Scene Segmentation of Colonoscopy Images

Figure 3

Examples of predictions for 4-class FCN8 model. Each subfigure represents a single frame, a ground truth annotation, and a prediction image. We use the following color-coding in the annotations: red for background (mucosa), blue for lumen, yellow for polyp, and green for specularity. (a), (b), (c), (d) show correct polyp segmentation, whereas (e), (d) show incorrect polyp segmentation.

(a)

(b)

(c)

(d)

(e)

(f)