28、计算机视觉技术实战：从ViT到SAM与FastSAM

计算机视觉技术实战：从ViT到SAM与FastSAM

1. Hugging Face ViT注意力分数可视化

在计算机视觉领域，理解模型对图像的关注点至关重要。Hugging Face的ViT（Vision Transformer）模型为我们提供了一种强大的工具来实现这一目标。以下是使用Python 3.9实现ViT注意力分数可视化的代码：

from PIL import Image 
import torch.nn as nn
import torch.nn.functional as F
import matplotlib.pyplot as plt
from transformers import ViTImageProcessor
from transformers import ViTModel

model_name = "facebook/dino-vits8" 
model = ViTModel.from_pretrained(model_name, add_pooling_layer=False) 
feature_extractor = ViTImageProcessor.from_pretrained(model_name, size=480) 
image = Image.open('d:/robin.jpg') 

# Using the pre-trained ViT model to get attention scores
input = feature_extractor(images=image, return_tensors='pt').pixel_values 
outputs = model(in