Table 3.

The comparative performance of MobileNet with Mish and ReLU using different optimizer methods.

Activation Function and Optimizer Methods	Accuracy %	Precision %	Recall %	F1 Measure %	Training Time
Nadam & ReLU	90.51	90.5	90.53	90.5	58 min 31 sec
Nadam & Mish	93.28	93.24	93.46	93.27	58 min 48 sec
Adamax & ReLU	87.17	87.31	86.9	87.04	58 min 05 sec
Adamax & Mish	90.44	90.42	90.64	90.43	58 min 18 sec
Adam & ReLU	86.91	87.19	86.56	86.75	58 min 27 sec
Adam & Mish	89.93	89.93	90.15	89.91	58 min 38 sec
Adadelta & ReLU	83.56	84.11	83.03	83.26	56 min 20 sec
Adadelta & Mish	87.6	87.54	87.73	87.57	56 min 34 sec
RMSprop & ReLU	83.04	83.68	82.47	82.71	55 min 47 sec
RMSprop & Mish	85.37	85.33	85.52	85.34	56 min 07 sec
Adagrad & ReLU	79.94	81.4	79.06	79.29	56 min 36 sec
Adagrad & Mish	80.89	80.8	80.94	80.84	57 min 03 sec
SGD & ReLU	74.48	75.93	73.48	73.52	59 min 27 sec
SGD & Mish	76.59	76.68	76.82	76.57	59 min 48 sec