Skip to main content

Table 1 The tiny VGG

From: Understanding deep learning via backtracking and deconvolution

ConvNet configuration
Input 224 × 224 × 3
Conv Receptive field Stride Padding Feature map
3 × 3 1 1 32
Conv Receptive field Stride Padding Feature map
3 × 3 1 1 32
Maxpool Filter Stride
2 × 2 2
Conv Receptive field Stride Padding Feature map
3 × 3 1 1 32
Conv Receptive field Stride Padding Feature map
3 × 3 1 1 32
Maxpool Filter Stride
2 × 2 2
Conv Receptive field Stride Padding Feature map
3 × 3 1 1 32
Conv Receptive field Stride Padding Feature map
3 × 3 1 1 32
Maxpool Filter Stride
2 × 2 2
FC 1024
FC 512
FC 256
FC No. of classes
Softmax