Table 1 Unsupervised model feature extraction network structure ResNet-50

From: A novel unsupervised contrastive learning framework for ancient Yi script character dataset construction

Layer Name

Output Size

Layer Details

Input

224 × 224 × 3

-

Conv1

112 × 112 × 64

7 × 7, 64 filters, stride 2

Max Pooling

56 × 56 × 64

3 × 3 max pool, stride 2

Conv2_x

56 × 56 × 256

[1 × 1, 64], [3 × 3, 64], [1 × 1, 256] × 3 blocks

Conv3_x

28 × 28 × 512

[1 × 1, 128], [3 × 3, 128], [1 × 1, 512] × 4 blocks

Conv4_x

14 × 14 × 1024

[1 × 1, 256], [3 × 3, 256], [1 × 1, 1024] × 6 blocks

Conv5_x

7 × 7 × 2048

[1 × 1, 512], [3 × 3, 512], [1 × 1, 2048] × 3 blocks

Average Pooling

1 × 1 × 2048

7 × 7 average pool

Fully Connected

1 × 1 × 128

128-d fully connected layer

Back to article page