EfficientNetV2 b0
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.
EfficientNetV2 b0
Classification, 224 × 224 input, 1,000 classes. Tensor sizes exclude batch.
LibreYOLO
EfficientNetV2 b0
Classification, 224 × 224 input, 1,000 classes. Tensor sizes exclude batch.
Network
Input
3 × 224 × 224
Pad SAME
L/R=0/1, T/B=0/1
Conv2d 3×3 / 2
32 × 112 × 112; p=0, g=1
BatchNorm2d
32 × 112 × 112
SiLU
32 × 112 × 112
Layer 1: ConvBnAct
16 × 112 × 112, repeat 1
Layer 2: EdgeResidual
32 × 56 × 56, repeat 1
Layer 2: EdgeResidual
32 × 56 × 56, repeat 1
Layer 3: EdgeResidual
48 × 28 × 28, repeat 1
Layer 3: EdgeResidual
48 × 28 × 28, repeat 1
Layer 4: InvertedResidual
96 × 14 × 14, repeat 1
Layer 4: InvertedResidual
96 × 14 × 14, repeat 2
Layer 5: InvertedResidual
112 × 14 × 14, repeat 1
Layer 5: InvertedResidual
112 × 14 × 14, repeat 4
Layer 6: InvertedResidual
192 × 7 × 7, repeat 1
Layer 6: InvertedResidual
192 × 7 × 7, repeat 7
Conv2d 1×1 / 1
1280 × 7 × 7; p=0, g=1
BatchNorm2d
1280 × 7 × 7
SiLU
1280 × 7 × 7
AdaptiveAvgPool2d
1280 × 1 × 1
Flatten
1280
Linear
1280 to 1000
Layer 1, unit 1
ConvBnAct
Input 32 × 112 × 112
Conv2d 3×3 / 1
16 × 112 × 112; p=1, g=1
BatchNorm2d
16 × 112 × 112
SiLU
16 × 112 × 112
Output 16 × 112 × 112
Repeat 1 times
Layer 2, unit 1
EdgeResidual
Input 16 × 112 × 112
Pad SAME
L/R=0/1, T/B=0/1
Conv2d 3×3 / 2
64 × 56 × 56; p=0, g=1
BatchNorm2d
64 × 56 × 56
SiLU
64 × 56 × 56
Conv2d 1×1 / 1
32 × 56 × 56; p=0, g=1
BatchNorm2d
32 × 56 × 56
Output 32 × 56 × 56
Repeat 1 times
Layer 2, unit 2
EdgeResidual
Input 32 × 56 × 56
Conv2d 3×3 / 1
128 × 56 × 56; p=1, g=1
BatchNorm2d
128 × 56 × 56
SiLU
128 × 56 × 56
Conv2d 1×1 / 1
32 × 56 × 56; p=0, g=1
BatchNorm2d
32 × 56 × 56
+
Output 32 × 56 × 56
Repeat 1 times
Layer 3, unit 1
EdgeResidual
Input 32 × 56 × 56
Pad SAME
L/R=0/1, T/B=0/1
Conv2d 3×3 / 2
128 × 28 × 28; p=0, g=1
BatchNorm2d
128 × 28 × 28
SiLU
128 × 28 × 28
Conv2d 1×1 / 1
48 × 28 × 28; p=0, g=1
BatchNorm2d
48 × 28 × 28
Output 48 × 28 × 28
Repeat 1 times
Layer 3, unit 2
EdgeResidual
Input 48 × 28 × 28
Conv2d 3×3 / 1
192 × 28 × 28; p=1, g=1
BatchNorm2d
192 × 28 × 28
SiLU
192 × 28 × 28
Conv2d 1×1 / 1
48 × 28 × 28; p=0, g=1
BatchNorm2d
48 × 28 × 28
+
Output 48 × 28 × 28
Repeat 1 times
Layer 4, unit 1
InvertedResidual
Input 48 × 28 × 28
Conv2d 1×1 / 1
192 × 28 × 28; p=0, g=1
BatchNorm2d
192 × 28 × 28
SiLU
192 × 28 × 28
Pad SAME
L/R=0/1, T/B=0/1
Conv2d 3×3 / 2
192 × 14 × 14; p=0, g=192
BatchNorm2d
192 × 14 × 14
SiLU
192 × 14 × 14
Spatial mean
192 × 1 × 1
Conv2d 1×1 / 1
12 × 1 × 1; p=0, g=1
SiLU
12 × 1 × 1
Conv2d 1×1 / 1
192 × 1 × 1; p=0, g=1
Sigmoid
192 × 1 × 1
Multiply gate with feature
192 × 14 × 14
Conv2d 1×1 / 1
96 × 14 × 14; p=0, g=1
BatchNorm2d
96 × 14 × 14
Output 96 × 14 × 14
Repeat 1 times
Layer 4, unit 2
InvertedResidual
Input 96 × 14 × 14
Conv2d 1×1 / 1
384 × 14 × 14; p=0, g=1
BatchNorm2d
384 × 14 × 14
SiLU
384 × 14 × 14
Conv2d 3×3 / 1
384 × 14 × 14; p=1, g=384
BatchNorm2d
384 × 14 × 14
SiLU
384 × 14 × 14
Spatial mean
384 × 1 × 1
Conv2d 1×1 / 1
24 × 1 × 1; p=0, g=1
SiLU
24 × 1 × 1
Conv2d 1×1 / 1
384 × 1 × 1; p=0, g=1
Sigmoid
384 × 1 × 1
Multiply gate with feature
384 × 14 × 14
Conv2d 1×1 / 1
96 × 14 × 14; p=0, g=1
BatchNorm2d
96 × 14 × 14
+
Output 96 × 14 × 14
Repeat 2 times
Layer 5, unit 1
InvertedResidual
Input 96 × 14 × 14
Conv2d 1×1 / 1
576 × 14 × 14; p=0, g=1
BatchNorm2d
576 × 14 × 14
SiLU
576 × 14 × 14
Conv2d 3×3 / 1
576 × 14 × 14; p=1, g=576
BatchNorm2d
576 × 14 × 14
SiLU
576 × 14 × 14
Spatial mean
576 × 1 × 1
Conv2d 1×1 / 1
24 × 1 × 1; p=0, g=1
SiLU
24 × 1 × 1
Conv2d 1×1 / 1
576 × 1 × 1; p=0, g=1
Sigmoid
576 × 1 × 1
Multiply gate with feature
576 × 14 × 14
Conv2d 1×1 / 1
112 × 14 × 14; p=0, g=1
BatchNorm2d
112 × 14 × 14
Output 112 × 14 × 14
Repeat 1 times
Layer 5, unit 2
InvertedResidual
Input 112 × 14 × 14
Conv2d 1×1 / 1
672 × 14 × 14; p=0, g=1
BatchNorm2d
672 × 14 × 14
SiLU
672 × 14 × 14
Conv2d 3×3 / 1
672 × 14 × 14; p=1, g=672
BatchNorm2d
672 × 14 × 14
SiLU
672 × 14 × 14
Spatial mean
672 × 1 × 1
Conv2d 1×1 / 1
28 × 1 × 1; p=0, g=1
SiLU
28 × 1 × 1
Conv2d 1×1 / 1
672 × 1 × 1; p=0, g=1
Sigmoid
672 × 1 × 1
Multiply gate with feature
672 × 14 × 14
Conv2d 1×1 / 1
112 × 14 × 14; p=0, g=1
BatchNorm2d
112 × 14 × 14
+
Output 112 × 14 × 14
Repeat 4 times
Layer 6, unit 1
InvertedResidual
Input 112 × 14 × 14
Conv2d 1×1 / 1
672 × 14 × 14; p=0, g=1
BatchNorm2d
672 × 14 × 14
SiLU
672 × 14 × 14
Pad SAME
L/R=0/1, T/B=0/1
Conv2d 3×3 / 2
672 × 7 × 7; p=0, g=672
BatchNorm2d
672 × 7 × 7
SiLU
672 × 7 × 7
Spatial mean
672 × 1 × 1
Conv2d 1×1 / 1
28 × 1 × 1; p=0, g=1
SiLU
28 × 1 × 1
Conv2d 1×1 / 1
672 × 1 × 1; p=0, g=1
Sigmoid
672 × 1 × 1
Multiply gate with feature
672 × 7 × 7
Conv2d 1×1 / 1
192 × 7 × 7; p=0, g=1
BatchNorm2d
192 × 7 × 7
Output 192 × 7 × 7
Repeat 1 times
Layer 6, unit 2
InvertedResidual
Input 192 × 7 × 7
Conv2d 1×1 / 1
1152 × 7 × 7; p=0, g=1
BatchNorm2d
1152 × 7 × 7
SiLU
1152 × 7 × 7
Conv2d 3×3 / 1
1152 × 7 × 7; p=1, g=1152
BatchNorm2d
1152 × 7 × 7
SiLU
1152 × 7 × 7
Spatial mean
1152 × 1 × 1
Conv2d 1×1 / 1
48 × 1 × 1; p=0, g=1
SiLU
48 × 1 × 1
Conv2d 1×1 / 1
1152 × 1 × 1; p=0, g=1
Sigmoid
1152 × 1 × 1
Multiply gate with feature
1152 × 7 × 7
Conv2d 1×1 / 1
192 × 7 × 7; p=0, g=1
BatchNorm2d
192 × 7 × 7
+
Output 192 × 7 × 7
Repeat 7 times
Repeated units are grouped only when operation parameters, tensor shapes and shortcut behavior match.
Source: libreyolo/models/efficientnetv2/nn.py. Revision a4d0ecc9e17f.
libreyolo.com