Table 5 ResNet-50 Layers.
From: Deep‑learning based osteoporosis classification in knee X‑rays using transfer‑learning approach
Layers | Description | Output Shape |
|---|---|---|
Conv1 | 7*7 convolution, stride 2, 64 filters | 112*112*64 |
Max Pooling | 3*3 max pooling, stride 2 | 56*56*64 |
Residual Block 1 (Stage 1) | Three Residual Blocks, each consisting of a 1*1,3*3 and 1*1 bottleneck structure, 64 filters | 56*56*256 |
Residual Block 2 (Stage 2) | 4 Residual Blocks, 128 filters | 28*28*512 |
Residual Block 3 (Stage 3) | 6 Residual Blocks, 256 filters | 14*14*1024 |
Residual Block 4 (Stage 4) | 3 Residual Blocks, 512 filters | 7*7*2048 |
Global Average Pooling | Averages each feature map into one value | 1*1*2048 |
Fully Connected Layers | Averages each feature map into one value | Number of classes (e.g., 1000 for ImageNet) |
