Table 2.

The architecture of the Resnet model. Each row represents a layer of the network, and the input of a particular layer is the output of the previous layer. Serial processes are represented as comma-separated parameters in each row. The number of times each Conv layer is repeated is prepended to each layer name. The feature map sizes are downsampled by a factor for 2 during the first iteration of Conv  3 to Conv 5. Resnet models with and without inclusion of the Auxiliary and dropout were constructed

Type	Patch size/strides	Input size
Conv 1	7 × 7/2	300 × 300 × 1
Max pool 1	3 × 3/2	147 × 147 × 64
3× Conv 2	1 × 1/1, 3 × 3/1, 1 × 1/1	74 × 74 × 64
4× Conv 3	1 × 1/1, 3 × 3/1, 1 × 1/1	74 × 74 × 256
23× Conv 4	1 × 1/1, 3 × 3/1, 1 × 1/1	37 × 37 × 512
Auxiliary	Avg pool 5 × 5/3, 1 × 1/1, linear, softmax	19 × 19 × 1024
3× Conv 5	1 × 1/1, 3 × 3/1, 1 × 1/1	19 × 19 × 1024
Avg pool	10 × 10/1	10 × 10 × 2048
Output	Dropout, linear, softmax	1 × 1 × 2048