A Multiorgan Segmentation Model for CT Volumes via Full Convolution-Deconvolution Network

Figure 4

Visualization of activation in our network. We select most representative activation maps for net visualization; (a) fuse1 layer (16 × 16), (b) new unpooling layer (32 × 32), (c) deconvolutional layer (32 × 32), (d) fuse2 layer (32 × 32), (e) new unpooling layer (64 × 64), (f) deconvolutional layer (64 × 64), (g) fuse3 layer (64 × 64), and (h) deconvolutional layer (512 × 512).

(a)

(b)

(c)

(d)

(e)

(f)

(g)

(h)