VGG-16
Fit diagram
Read at 100%
Clear selection
Download SVG
Download PNG
Click a block to read its description, or select it with Tab and Enter.
VGG-16
Classification, 224 × 224 input, 1,000 classes. Tensor sizes exclude batch.
LibreYOLO
VGG-16
Classification, 224 × 224 input, 1,000 classes. Tensor sizes exclude batch.
Stage 1
Input
3 × 224 × 224
Conv2d 3×3 / 1
64 × 224 × 224; p=1, g=1
ReLU
64 × 224 × 224
Conv2d 3×3 / 1
64 × 224 × 224; p=1, g=1
ReLU
64 × 224 × 224
MaxPool2d
64 × 112 × 112
Stage 2
Input
64 × 112 × 112
Conv2d 3×3 / 1
128 × 112 × 112; p=1, g=1
ReLU
128 × 112 × 112
Conv2d 3×3 / 1
128 × 112 × 112; p=1, g=1
ReLU
128 × 112 × 112
MaxPool2d
128 × 56 × 56
Stage 3
Input
128 × 56 × 56
Conv2d 3×3 / 1
256 × 56 × 56; p=1, g=1
ReLU
256 × 56 × 56
Conv2d 3×3 / 1
256 × 56 × 56; p=1, g=1
ReLU
256 × 56 × 56
Conv2d 3×3 / 1
256 × 56 × 56; p=1, g=1
ReLU
256 × 56 × 56
MaxPool2d
256 × 28 × 28
Stage 4
Input
256 × 28 × 28
Conv2d 3×3 / 1
512 × 28 × 28; p=1, g=1
ReLU
512 × 28 × 28
Conv2d 3×3 / 1
512 × 28 × 28; p=1, g=1
ReLU
512 × 28 × 28
Conv2d 3×3 / 1
512 × 28 × 28; p=1, g=1
ReLU
512 × 28 × 28
MaxPool2d
512 × 14 × 14
Stage 5
Input
512 × 14 × 14
Conv2d 3×3 / 1
512 × 14 × 14; p=1, g=1
ReLU
512 × 14 × 14
Conv2d 3×3 / 1
512 × 14 × 14; p=1, g=1
ReLU
512 × 14 × 14
Conv2d 3×3 / 1
512 × 14 × 14; p=1, g=1
ReLU
512 × 14 × 14
MaxPool2d
512 × 7 × 7
Classifier
AdaptiveAvgPool2d
512 × 7 × 7
Flatten
25,088
Linear
25088 to 4096
ReLU
4096
Dropout (eval identity)
p=0.5; 4096
Linear
4096 to 4096
ReLU
4096
Dropout (eval identity)
p=0.5; 4096
Linear
4096 to 1000
Continue at stage 2
Continue at stage 3
Continue at stage 4
Continue at stage 5
Continue at classifier
Conv2d: bias=True, 3×3, stride 1, padding 1. MaxPool2d: 2×2, stride 2, padding 0.
Stage inputs are the preceding stage outputs. Dropout is identity in eval. Output: 1,000 logits.
Source: libreyolo/models/vgg/nn.py. Revision a4d0ecc9e17f.
libreyolo.com