数据集:

• imagenet-1k 库名称: timm 标签:
• 图像分类
• timm
• rdnet 许可证: bsd-3-clause

rdnet_tiny.nv_in1k模型卡

一个RDNet图像分类模型。在ImageNet-1k上训练，原始torchvision权重。

模型详情

• 模型类型: 图像分类/特征骨干网络
• 模型统计:
  • ImageNet-1k验证集top-1准确率: 82.8%
  • 参数量(M): 24
  • GMACs: 5.0
  • 图像尺寸: 224 x 224
• 论文:
  • DenseNets重装上阵：超越ResNets和ViTs的范式转变: https://arxiv.org/abs/2403.19588
• 数据集: ImageNet-1k

模型使用

图像分类

from urllib.request import urlopen
from PIL import Image
import timm
import torch

img = Image.open(urlopen(
    'https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/beignets-task-guide.png'
))

model = timm.create_model('rdnet_tiny.nv_in1k', pretrained=True)
model = model.eval()

# 获取模型特定的变换(归一化、调整大小)
data_config = timm.data.resolve_model_data_config(model)
transforms = timm.data.create_transform(**data_config, is_training=False)

output = model(transforms(img).unsqueeze(0))  # 将单张图像扩展为批量大小为1

top5_probabilities, top5_class_indices = torch.topk(output.softmax(dim=1) * 100, k=5)

特征图提取

from urllib.request import urlopen
from PIL import Image
import timm

img = Image.open(urlopen(
    'https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/beignets-task-guide.png'
))

model = timm.create_model(
    'rdnet_tiny.nv_in1k',
    pretrained=True,
    features_only=True,
)
model = model.eval()

# 获取模型特定的变换(归一化、调整大小)
data_config = timm.data.resolve_model_data_config(model)
transforms = timm.data.create_transform(**data_config, is_training=False)

output = model(transforms(img).unsqueeze(0))  # 将单张图像扩展为批量大小为1

for o in output:
    # 打印输出中每个特征图的形状
    # 例如:
    # torch.Size([1, 256, 56, 56])
    # torch.Size([1, 440, 28, 28])
    # torch.Size([1, 744, 14, 14])
    # torch.Size([1, 1040, 7, 7])

    print(o.shape)

图像嵌入

from urllib.request import urlopen
from PIL import Image
import timm

img = Image.open(urlopen(
    'https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/beignets-task-guide.png'
))

model = timm.create_model(
    'rdnet_tiny.nv_in1k',
    pretrained=True,
    num_classes=0,  # 移除分类器nn.Linear
)
model = model.eval()

# 获取模型特定的变换(归一化、调整大小)
data_config = timm.data.resolve_model_data_config(model)
transforms = timm.data.create_transform(**data_config, is_training=False)

output = model(transforms(img).unsqueeze(0))  # 输出是(batch_size, num_features)形状的张量

# 或等效地(无需设置num_classes=0)

output = model.forward_features(transforms(img).unsqueeze(0))
# 输出未池化，是(1, 1040, 7, 7)形状的张量

output = model.forward_head(output, pre_logits=True)
# 输出是(1, num_features)形状的张量

引用

@misc{kim2024densenets,
    title={DenseNets重装上阵：超越ResNets和ViTs的范式转变}, 
    author={Donghyun Kim and Byeongho Heo and Dongyoon Han},
    year={2024},
    eprint={2403.19588},
    archivePrefix={arXiv},
}