Large-Scale Video Retrieval via Deep Local Convolutional Features

Table 1

Structure of VGG16.

Layer	Output size
Conv3-64	224  224  64
Conv3-64	224  224  64
Max-pooling	112  112  64
Conv3-128	112  112  128
Conv3-128	112  112  128
Max-pooling	56  56  128
Conv3-256	56  56  256
Conv3-256	56  56  256
Conv3-256	56  56  256
Max-pooling	28  28  256
Conv3-512	28  28  512
Conv3-512	28  28  512
Conv3-512	28  28  512
Max-pooling	14  14  512
Conv3-512	14  14  512
Conv3-512	14  14  512
Conv3-512	14  14  512
Max-pooling	7  7  512