Table 4: Classification accuracies with different parameters during fine-tuning of the ResNet. The numbers inside parenthesis indicate batch size and number of iterations.

Networkslabel1label2label3label4label5label6label7label8label9overall

ResNet (16,2496)90.91%88.89%100%100%96.88%100%90.28%88.20%97.55%94.98%
ResNet (16,4992)98.1898.7710098.6296.8810096.5388.8298.7797.19
ResNet (16,9984)98.18%97.53%100%98.62%96.88%100%97.22%86.96%98.77%96.94%
ResNet (32,2496)97.27%95.06%100%97.93%96.88%100%95.14%86.34%99.39%96.34%
ResNet (32,4992)97.27%95.06%100%97.24%96.88%100%96.53%86.96%99.39%96.51%
ResNet (32,9984)96.36%96.30%100%99.31%96.88%100%94.44%88.20%98.77%96.60%
ResNet (64,2496)93.64%93.83%100%97.24%96.88%100%94.44%87.58%99.39%95.92%
ResNet (64,4992)94.55%95.29%100%96.55%95.83%100%95.83%86.96%99.39%95.83%
ResNet (64,9984)95.45%93.83%100%97.93%96.88%100%94.44%87.58%99.39%96.17%