VGG

2024-04-08 10:51| 来源: 网络整理| 查看: 265

VGG

pytorch / vision Last updated on Feb 12, 2021 VGG-11

Parameters 133 Million FLOPs 8 Billion File Size 506.84 MB Training Data ImageNet Training Resources 8x NVIDIA V100 GPUs Training Time Training Techniques Weight Decay, SGD with Momentum Architecture Convolution, Dropout, Dense Connections, ReLU, Max Pooling, Softmax ID vgg11 LR 0.2 Epochs 90 LR Gamma 0.1 Momentum 0.9 Batch Size 32 LR Step Size 30 Weight Decay 0.0001 SHOW MORE SHOW LESS

VGG-11 with batch normalization

Parameters 133 Million FLOPs 8 Billion File Size 506.88 MB Training Data ImageNet Training Resources 8x NVIDIA V100 GPUs Training Time Training Techniques Weight Decay, SGD with Momentum Architecture Batch Normalization, Convolution, Dropout, Dense Connections, ReLU, Max Pooling, Softmax ID vgg11_bn LR 0.2 Epochs 90 LR Gamma 0.1 Momentum 0.9 Batch Size 32 LR Step Size 30 Weight Decay 0.0001 SHOW MORE SHOW LESS

VGG-13

Parameters 133 Million FLOPs 11 Billion File Size 507.54 MB Training Data ImageNet Training Resources 8x NVIDIA V100 GPUs Training Time Training Techniques Weight Decay, SGD with Momentum Architecture Convolution, Dropout, Dense Connections, ReLU, Max Pooling, Softmax ID vgg13 LR 0.2 Epochs 90 LR Gamma 0.1 Momentum 0.9 Batch Size 32 LR Step Size 30 Weight Decay 0.0001 SHOW MORE SHOW LESS

VGG-13 with batch normalization

Parameters 133 Million FLOPs 11 Billion File Size 507.59 MB Training Data ImageNet Training Resources 8x NVIDIA V100 GPUs Training Time Training Techniques Weight Decay, SGD with Momentum Architecture Batch Normalization, Convolution, Dropout, Dense Connections, ReLU, Max Pooling, Softmax ID vgg13_bn LR 0.2 Epochs 90 LR Gamma 0.1 Momentum 0.9 Batch Size 32 LR Step Size 30 Weight Decay 0.0001 SHOW MORE SHOW LESS

VGG-16

Parameters 138 Million FLOPs 15 Billion File Size 527.80 MB Training Data ImageNet Training Resources 8x NVIDIA V100 GPUs Training Time Training Techniques Weight Decay, SGD with Momentum Architecture Convolution, Dropout, Dense Connections, ReLU, Max Pooling, Softmax ID vgg16 LR 0.2 Epochs 90 LR Gamma 0.1 Momentum 0.9 Batch Size 32 LR Step Size 30 Weight Decay 0.0001 SHOW MORE SHOW LESS

VGG-16 with batch normalization

Parameters 138 Million FLOPs 16 Billion File Size 527.87 MB Training Data ImageNet Training Resources 8x NVIDIA V100 GPUs Training Time Training Techniques Weight Decay, SGD with Momentum Architecture Batch Normalization, Convolution, Dropout, Dense Connections, ReLU, Max Pooling, Softmax ID vgg16_bn LR 0.2 Epochs 90 LR Gamma 0.1 Momentum 0.9 Batch Size 32 LR Step Size 30 Weight Decay 0.0001 SHOW MORE SHOW LESS

VGG-19

Parameters 144 Million FLOPs 20 Billion File Size 548.05 MB Training Data ImageNet Training Resources 8x NVIDIA V100 GPUs Training Time Training Techniques Weight Decay, SGD with Momentum Architecture Convolution, Dropout, Dense Connections, ReLU, Max Pooling, Softmax ID vgg19 LR 0.2 Epochs 90 LR Gamma 0.1 Momentum 0.9 Batch Size 32 LR Step Size 30 Weight Decay 0.0001 SHOW MORE SHOW LESS

VGG-19 with batch normalization

Parameters 144 Million FLOPs 20 Billion File Size 548.14 MB Training Data ImageNet Training Resources 8x NVIDIA V100 GPUs Training Time Training Techniques Weight Decay, SGD with Momentum Architecture Batch Normalization, Convolution, Dropout, Dense Connections, ReLU, Max Pooling, Softmax ID vgg19_bn LR 0.2 Epochs 90 LR Gamma 0.1 Momentum 0.9 Batch Size 32 LR Step Size 30 Weight Decay 0.0001 SHOW MORE SHOW LESS

README.md Summary

VGG is a classical convolutional neural network architecture. It was based on an analysis of how to increase the depth of such networks. The network utilises small 3 x 3 filters. Otherwise the network is characterized by its simplicity: the only other components being pooling layers and a fully connected layer.

How do I load this model?

To load a pretrained model:

import torchvision.models as models vgg16 = models.vgg16(pretrained=True)

Replace the model name with the variant you want to use, e.g. vgg16. You can find the IDs in the model summaries at the top of this page.

To evaluate the model, use the image classification recipes from the library.

python train.py --test-only --model='' How do I train this model?

You can follow the torchvision recipe on GitHub for training a new model afresh.

Citation @InProceedings{Simonyan15, author = "Karen Simonyan and Andrew Zisserman", title = "Very Deep Convolutional Networks for Large-Scale Image Recognition", booktitle = "International Conference on Learning Representations", year = "2015", } Results Image Classification on ImageNet

Image Classification on ImageNet MODEL TOP 1 ACCURACY TOP 5 ACCURACY VGG-19 with batch normalization 74.24% 91.85% VGG-16 with batch normalization 73.37% 91.5% VGG-19 72.38% 90.88% VGG-16 71.59% 90.38% VGG-13 with batch normalization 71.55% 90.37% VGG-11 with batch normalization 70.38% 89.81% VGG-13 69.93% 89.25% VGG-11 69.02% 88.63%

【本文地址】

公司简介

联系我们