monetjoe/cv_backbones

该数据集整合了PyTorch官方网站上提供的预训练计算机视觉模型的主干网络，主要包括在ImageNet1K数据集上预训练的各种卷积神经网络（CNNs）和视觉Transformer模型。数据集分为V1和V2两个子集，涵盖了多个经典和先进的视觉模型版本。这些预训练的主干网络为用户在图像识别、目标检测和图像分割等任务中的迁移学习提供了坚实的基础，同时也为研究人员和从业者提供了在不同场景中灵活应用这些预训练模型的便利选择。

This repository consolidates the collection of backbone networks for pre-trained computer vision models available on the PyTorch official website. It mainly includes various Convolutional Neural Networks (CNNs) and Vision Transformer models pre-trained on the ImageNet1K dataset. The entire collection is divided into two subsets, V1 and V2, encompassing multiple classic and advanced versions of visual models. These pre-trained backbone networks provide users with a robust foundation for transfer learning in tasks such as image recognition, object detection, and image segmentation. Simultaneously, it offers a convenient choice for researchers and practitioners to flexibly apply these pre-trained models in different scenarios.

原始信息汇总

数据集概述

基本信息

• 许可证: MIT
• 任务类别:
  • 图像分类
  • 特征提取
• 语言: 英语
• 标签: 代码
• 名称: Vi-Backbones
• 大小类别: n<1K

数据集描述

该数据集整合了PyTorch官方网站上可用的预训练计算机视觉模型骨干网络。主要包含在ImageNet1K数据集上预训练的各种卷积神经网络（CNN）和Vision Transformer模型。整个集合分为两个子集，V1和V2，涵盖了多个经典和高级版本的视觉模型。这些预训练的骨干网络为用户提供了在图像识别、目标检测和图像分割等任务中进行迁移学习的坚实基础。同时，它为研究人员和从业者提供了灵活应用这些预训练模型在不同场景中的便利选择。

数据字段

字段名称	字段类型	输入图像尺寸	预训练模型文件URL
backbone name	backbone type	input image size	url of pretrained model .pth file

参数统计

IMAGENET1K_V1

骨干网络名称	参数数量(M)
SqueezeNet1_0	1.2
SqueezeNet1_1	1.2
ShuffleNet_V2_X0_5	1.4
MNASNet0_5	2.2
ShuffleNet_V2_X1_0	2.3
MobileNet_V3_Small	2.5
MNASNet0_75	3.2
MobileNet_V2	3.5
ShuffleNet_V2_X1_5	3.5
RegNet_Y_400MF	4.3
MNASNet1_0	4.4
EfficientNet_B0	5.3
MobileNet_V3_Large	5.5
RegNet_X_400MF	5.5
MNASNet1_3	6.3
RegNet_Y_800MF	6.4
GoogLeNet	6.6
RegNet_X_800MF	7.3
ShuffleNet_V2_X2_0	7.4
EfficientNet_B1	7.8
DenseNet121	8
EfficientNet_B2	9.1
RegNet_X_1_6GF	9.2
RegNet_Y_1_6GF	11.2
ResNet18	11.7
EfficientNet_B3	12.2
DenseNet169	14.1
RegNet_X_3_2GF	15.3
EfficientNet_B4	19.3
RegNet_Y_3_2GF	19.4
DenseNet201	20
EfficientNet_V2_S	21.5
ResNet34	21.8
ResNeXt50_32X4D	25
ResNet50	25.6
Inception_V3	27.2
Swin_T	28.3
Swin_V2_T	28.4
ConvNeXt_Tiny	28.6
DenseNet161	28.7
EfficientNet_B5	30.4
MaxVit_T	30.9
RegNet_Y_8GF	39.4
RegNet_X_8GF	39.6
EfficientNet_B6	43
ResNet101	44.5
Swin_S	49.6
Swin_V2_S	49.7
ConvNeXt_Small	50.2
EfficientNet_V2_M	54.1
RegNet_X_16GF	54.3
ResNet152	60.2
AlexNet	61.1
EfficientNet_B7	66.3
Wide_ResNet50_2	68.9
ResNeXt101_64X4D	83.5
RegNet_Y_16GF	83.6
ViT_B_16	86.6
Swin_B	87.8
Swin_V2_B	87.9
ViT_B_32	88.2
ConvNeXt_Base	88.6
ResNeXt101_32X8D	88.8
RegNet_X_32GF	107.8
EfficientNet_V2_L	118.5
Wide_ResNet101_2	126.9
VGG11_BN	132.9
VGG11	132.9
VGG13	133
VGG13_BN	133.1
VGG16_BN	138.4
VGG16	138.4
VGG19_BN	143.7
VGG19	143.7
RegNet_Y_32GF	145
ConvNeXt_Large	197.8
ViT_L_16	304.3
ViT_L_32	306.5

IMAGENET1K_V2

骨干网络名称	参数数量(M)
MobileNet_V2	3.5
RegNet_Y_400MF	4.3
MobileNet_V3_Large	5.5
RegNet_X_400MF	5.5
RegNet_Y_800MF	6.4
RegNet_X_800MF	7.3
EfficientNet_B1	7.8
RegNet_X_1_6GF	9.2
RegNet_Y_1_6GF	11.2
RegNet_X_3_2GF	15.3
RegNet_Y_3_2GF	19.4
ResNeXt50_32X4D	25
ResNet50	25.6
RegNet_Y_8GF	39.4
RegNet_X_8GF	39.6
ResNet101	44.5
RegNet_X_16GF	54.3
ResNet152	60.2
Wide_ResNet50_2	68.9
RegNet_Y_16GF	83.6
ResNeXt101_32X8D	88.8
RegNet_X_32GF	107.8
Wide_ResNet101_2	126.9
RegNet_Y_32GF	145