A Deep Learning Model for Quick and Accurate Rock Recognition with Smartphones

Figure 10

Model accuracy versus training epoch on various batch sizes: (a) result on the MobileNet model; (b, c) result on SqueezeNet and ShuffleNet models, respectively. Orange, yellow, green, and wine curves represent batch sizes equal 8, 16, 32, and 48, respectively.

(a)

(b)

(c)