maxvit_xlarge_tf_224.in21k模型卡片

一个官方的MaxViT图像分类模型。由论文作者在ImageNet-21k（ImageNet-22k的21843个Google特定实例）上使用tensorflow训练而成。

由Ross Wightman将官方的Tensorflow实现（ https://github.com/google-research/maxvit ）移植到PyTorch。

maxxvit.py 中的模型变体

MaxxViT包括一些相关的模型架构，它们共享一个共同的结构，包括：

• CoAtNet-在早期阶段使用MBConv（深度可分离）卷积块与后期的自注意力Transformer块相结合。
• MaxViT-统一的块分布在所有阶段，每个阶段包含一个MBConv（深度可分离）卷积块，后面是两个不同分区方案（窗口跟随网格）的自注意力块。
• CoAtNeXt-使用ConvNeXt块代替CoAtNet中的MBConv块的timm特定架构。所有标准化层都是LayerNorm（没有BatchNorm）。
• MaxxViT-使用ConvNeXt块代替MaxViT中的MBConv块的timm特定架构。所有标准化层都是LayerNorm（没有BatchNorm）。
• MaxxViT-V2-MaxxViT的变体，删除窗口块注意力，仅保留ConvNeXt块和网格注意力，通过增加宽度来补偿。

除了上述主要变体外，模型之间还有更细微的差异。所有带有字符串rw的模型名称都是timm特定配置，具有有利于PyTorch eager use的建模调整。这些模型是在进行初始的模型复现训练时创建的，因此存在差异。所有带有字符串tf的模型都是与原始论文作者的Tensorflow模型完全匹配，并将权重移植到PyTorch。这包括一些MaxViT模型。官方的CoAtNet模型从未发布。

模型详细信息

• 模型类型：图像分类/特征主干
• 模型统计信息：
  • 参数（M）：507.0
  • GMACs：97.5
  • 激活数（M）：191.0
  • 图像大小：224 x 224
• 论文：
  • MaxViT: Multi-Axis Vision Transformer： https://arxiv.org/abs/2204.01697
• 数据集：ImageNet-21k

模型用途

图像分类

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('maxvit_xlarge_tf_224.in21k', pretrained=True)
model = model.eval()

# get model specific transforms (normalization, resize)
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))  # unsqueeze single image into batch of 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(
    'maxvit_xlarge_tf_224.in21k',
    pretrained=True,
    features_only=True,
)
model = model.eval()

# get model specific transforms (normalization, resize)
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))  # unsqueeze single image into batch of 1

for o in output:
    # print shape of each feature map in output
    # e.g.:
    #  torch.Size([1, 192, 112, 112])
    #  torch.Size([1, 192, 56, 56])
    #  torch.Size([1, 384, 28, 28])
    #  torch.Size([1, 768, 14, 14])
    #  torch.Size([1, 1536, 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(
    'maxvit_xlarge_tf_224.in21k',
    pretrained=True,
    num_classes=0,  # remove classifier nn.Linear
)
model = model.eval()

# get model specific transforms (normalization, resize)
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))  # output is (batch_size, num_features) shaped tensor

# or equivalently (without needing to set num_classes=0)

output = model.forward_features(transforms(img).unsqueeze(0))
# output is unpooled, a (1, 1536, 7, 7) shaped tensor

output = model.forward_head(output, pre_logits=True)
# output is a (1, num_features) shaped tensor

模型对比

按Top-1指标

按吞吐量（样本/秒）

引用

@misc{rw2019timm,
  author = {Ross Wightman},
  title = {PyTorch Image Models},
  year = {2019},
  publisher = {GitHub},
  journal = {GitHub repository},
  doi = {10.5281/zenodo.4414861},
  howpublished = {\url{https://github.com/huggingface/pytorch-image-models}}
}

@article{tu2022maxvit,
  title={MaxViT: Multi-Axis Vision Transformer},
  author={Tu, Zhengzhong and Talebi, Hossein and Zhang, Han and Yang, Feng and Milanfar, Peyman and Bovik, Alan and Li, Yinxiao},
  journal={ECCV},
  year={2022},
}        

@article{dai2021coatnet,
  title={CoAtNet: Marrying Convolution and Attention for All Data Sizes},
  author={Dai, Zihang and Liu, Hanxiao and Le, Quoc V and Tan, Mingxing},
  journal={arXiv preprint arXiv:2106.04803},
  year={2021}
}