Table 3.

Evaluation metrics of the different neural network architectures in the TCGA melanoma test dataset.

Network	Accuracy (%)	Precision (%)	Recall (%)	Parameters (in millions)
InceptionV3	97.5	94.2	96.7	~22.4
InceptionResNetV2	97.7	94.1	97.3	~54.8
ResNet50	93.8	92.2	88.9	~24.2
Xception	98.8	96.9	98.5	~21.4
Our custom-made CNN	96.7	93.6	93.6	~1.9

The bold values in the Accuracy (%), Precision (%) and Recall (%) fields indicate the highest value i.e. the best performance achieved amongst the networks under comparison. The bold value in the Parameters (in millions) field indicate the network with the fewer parameters used amongst the networks under comparison.