PyTorch-模型验证

news2024/10/7 18:23:34

目的：利用已经训练好的网络模型，输入数据进行测试。（相当于将其运用于真实场景中）

Cifar-10分类：

test1：放入一张狗狗的图片

进行验证：test.py

import torch
import torchvision
from PIL import Image
from torch import nn
from torch.nn import Conv2d, MaxPool2d, Flatten, Linear

img_path = './imgs/1.png'
img = Image.open(img_path)
# print(img)  # <PIL.PngImagePlugin.PngImageFile image mode=RGBA size=153x153 at 0x24C0AA94820>
# 注意：png是四个通道,除了RGB外还有一个透明通道。
img = img.convert('RGB')
transform = torchvision.transforms.Compose([
    torchvision.transforms.Resize((32, 32)),
    torchvision.transforms.ToTensor()
])
img = transform(img)
# print(img.shape)  # torch.Size([3, 32, 32])
model = torch.load('./model_save/myModule_0.pth')
print(model)
img = torch.reshape(img, (1, 3, 32, 32))
model.eval()
with torch.no_grad():
    output = model(img)
print(output)
print(output.argmax(1))

MyModule(
(model): Sequential(
(0): Conv2d(3, 32, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2))
(1): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
(2): Conv2d(32, 32, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2))
(3): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
(4): Conv2d(32, 64, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2))
(5): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
(6): Flatten(start_dim=1, end_dim=-1)
(7): Linear(in_features=1024, out_features=64, bias=True)
(8): Linear(in_features=64, out_features=10, bias=True)
)
)
tensor([[-1.8267, -0.0232, 0.6100, 0.9207, 0.7240, 1.2317, 0.8643, 0.2022,
-2.3572, -1.3339]])
tensor([5])

Process finished with exit code 0

预测结果：正确

test2：放入一张飞机的图片

tensor([[ 0.1496, 0.5647, 0.4304, 0.0127, -0.0547, -0.0984, -0.3994, 0.4792,
-0.1503, 0.4344]])
tensor([1])

预测结果：错误

因为我们加载的是“myModule_0.pth”，还没有进行太多轮训练。

可以使用训练过多轮后的模型，再进行加载。

test.py

import torch
import torchvision
from PIL import Image
from torch import nn
from torch.nn import Conv2d, MaxPool2d, Flatten, Linear

img_path = './imgs/2.png'
img = Image.open(img_path)
# print(img)  # <PIL.PngImagePlugin.PngImageFile image mode=RGBA size=153x153 at 0x24C0AA94820>
# 注意：png是四个通道,除了RGB外还有一个透明通道。
img = img.convert('RGB')
transform = torchvision.transforms.Compose([
    torchvision.transforms.Resize((32, 32)),
    torchvision.transforms.ToTensor()
])
img = transform(img)
# print(img.shape)  # torch.Size([3, 32, 32])
model = torch.load('./model_save/myModule_9.pth')
print(model)
img = torch.reshape(img, (1, 3, 32, 32))
model.eval()
with torch.no_grad():
    output = model(img)
print(output)
print(output.argmax(1))