原创:余晓龙
Pytorch中提供了很多已经在ImageNet数据集上训练好的模型了,可以直接被加载到模型中进行预测任务。预训练模型存放在Pytorch的torchvision中库,在torchvision库的models模块下可以查看内置的模型,models模块中的模型包含四大类,如图所示:
一、图像分类代码实现
# coding: utf-8
from PIL import Image
import matplotlib.pyplot as plt
plt.rcParams['font.sans-serif'] = ['SimHei'] # 步骤一(替换sans-serif字体)
plt.rcParams['axes.unicode_minus'] = False # 步骤二(解决坐标轴负数的负号显示问题)
import json
import numpy as np
import torch
import torch.nn.functional as F
from torchvision import models, transforms
# 1.下载并加载预训练模型
model = models.resnet18(pretrained=True)
model = model.eval()
# 2.加载标签并对输入数据进行处理
labels_path = './imagenet_class_index.json'
with open(labels_path) as json_data:
idx2labels = json.load(json_data)
# print(idx2labels)
def getone(onestr):
return onestr.replace(',', ' ')
# 加载中文标签
with open('./zh_label.csv', 'r+', encoding='gbk') as f:
# print(f)
# print(map(getone, list(f)))
zh_labels = list(map(getone, list(f)))
print(len(zh_labels), type(zh_labels), zh_labels[:5])
transform = transforms.Compose([
transforms.Resize(256),
transforms.CenterCrop(224),
transforms.ToTensor(),
transforms.Normalize(
mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225]
)
])
# 3.使用模型进行预测
def preimg(img):
if img.mode == 'RGBA':
ch = 4
a = np.asarray(img)[:, :, :3]
img = Image.fromarray(a)
return img
im = preimg(Image.open('panda.jpg'))
transformed_img = transform(im)
inputimg = transformed_img.unsqueeze(0)
output = model(inputimg)
output = F.softmax(output, dim=1)
prediction_score, pred_label_i