update for change of exp2 and add exp3
This commit is contained in:
114
exp2/modified/mnist.py
Normal file
114
exp2/modified/mnist.py
Normal file
@@ -0,0 +1,114 @@
|
||||
from pathlib import Path
|
||||
import numpy
|
||||
import torch
|
||||
from torch.utils.data import DataLoader, Dataset
|
||||
from torchvision.transforms import v2 as tvtrans
|
||||
from torchvision import datasets
|
||||
import torch.nn.functional as F
|
||||
|
||||
|
||||
class CNN(torch.nn.Module):
|
||||
"""卷积神经网络模型"""
|
||||
|
||||
def __init__(self):
|
||||
super(CNN, self).__init__()
|
||||
|
||||
self.conv1 = torch.nn.Conv2d(1, 32, kernel_size=(3, 3))
|
||||
self.pool1 = torch.nn.MaxPool2d(kernel_size=(2, 2))
|
||||
self.conv2 = torch.nn.Conv2d(32, 64, kernel_size=(3, 3))
|
||||
self.pool2 = torch.nn.MaxPool2d(kernel_size=(2, 2))
|
||||
self.conv3 = torch.nn.Conv2d(64, 64, kernel_size=(3, 3))
|
||||
self.flatten = torch.nn.Flatten()
|
||||
# 28x28过第一轮卷积后变为26x26，过第一轮池化后变为13x13
|
||||
# 过第二轮卷积后变为11x11，过第二轮池化后变为5x5
|
||||
# 过第三轮卷积后变为3x3。
|
||||
# 最后一轮卷积核个数为64。
|
||||
self.fc1 = torch.nn.Linear(64 * 3 * 3, 64)
|
||||
torch.nn.init.xavier_normal_(self.fc1.weight)
|
||||
torch.nn.init.zeros_(self.fc1.bias)
|
||||
self.fc2 = torch.nn.Linear(64, 10)
|
||||
torch.nn.init.xavier_normal_(self.fc2.weight)
|
||||
torch.nn.init.zeros_(self.fc2.bias)
|
||||
|
||||
def forward(self, x):
|
||||
x = F.relu(self.conv1(x))
|
||||
x = self.pool1(x)
|
||||
x = F.relu(self.conv2(x))
|
||||
x = self.pool2(x)
|
||||
x = F.relu(self.conv3(x))
|
||||
x = self.flatten(x)
|
||||
x = F.relu(self.fc1(x))
|
||||
x = self.fc2(x)
|
||||
return F.softmax(x, dim=1)
|
||||
|
||||
|
||||
class MnistDataset(Dataset):
|
||||
"""用于加载Mnist的自定义数据集"""
|
||||
|
||||
shape: int
|
||||
transform: tvtrans.Transform
|
||||
images_data: numpy.ndarray
|
||||
labels_data: torch.Tensor
|
||||
|
||||
def __init__(self, images: numpy.ndarray, labels: numpy.ndarray, transform: tvtrans.Transform):
|
||||
images_len: int = images.shape[0]
|
||||
labels_len: int = labels.shape[0]
|
||||
assert (images_len == labels_len)
|
||||
self.shape = images_len
|
||||
|
||||
self.images_data = images
|
||||
self.labels_data = torch.from_numpy(labels)
|
||||
self.transform = transform
|
||||
|
||||
def __getitem__(self, index):
|
||||
return self.transform(self.images_data[index]), self.labels_data[index]
|
||||
|
||||
def __len__(self):
|
||||
return self.shape
|
||||
|
||||
|
||||
class MnistDataSource:
|
||||
"""用于读取MNIST数据的数据读取器"""
|
||||
|
||||
train_loader: DataLoader
|
||||
test_loader: DataLoader
|
||||
|
||||
def __init__(self, batch_size: int):
|
||||
dataset_path = Path(__file__).resolve().parent.parent / 'datasets' / 'mnist.npz'
|
||||
dataset = numpy.load(dataset_path)
|
||||
|
||||
# 所有图片均为黑底白字
|
||||
# 6万张训练图片：60000x28x28。标签只有第一维。
|
||||
train_images: numpy.ndarray = dataset['x_train']
|
||||
train_labels: numpy.ndarray = dataset['y_train']
|
||||
# 1万张测试图片：10000x28x28。标签只有第一维。
|
||||
test_images: numpy.ndarray = dataset['x_test']
|
||||
test_labels: numpy.ndarray = dataset['y_test']
|
||||
|
||||
# 定义数据转换器
|
||||
trans = tvtrans.Compose([
|
||||
# 从uint8转换为float32并自动归一化到0-1区间
|
||||
# tvtrans.ToTensor(),
|
||||
tvtrans.ToImage(),
|
||||
tvtrans.ToDtype(torch.float32, scale=True),
|
||||
|
||||
# 为了符合后面图像的输入颜色通道条件，要在最后挤出一个新的维度
|
||||
#tvtrans.Lambda(lambda x: x.unsqueeze(-1))
|
||||
|
||||
# 这个特定的标准化参数 (0.1307, 0.3081) 是 MNIST 数据集的标准化参数，这些数值是MNIST训练集的全局均值和标准差。
|
||||
# 这种标准化有助于模型训练时的数值稳定性和收敛速度。
|
||||
#tvtrans.Normalize((0.1307,), (0.3081,)),
|
||||
])
|
||||
|
||||
# 创建数据集
|
||||
train_dataset = MnistDataset(train_images, train_labels, transform=trans)
|
||||
test_dataset = MnistDataset(test_images, test_labels, transform=trans)
|
||||
|
||||
|
||||
# 赋值到自身
|
||||
self.train_loader = DataLoader(dataset=train_dataset,
|
||||
batch_size=batch_size,
|
||||
shuffle=False)
|
||||
self.test_loader = DataLoader(dataset=test_dataset,
|
||||
batch_size=batch_size,
|
||||
shuffle=False)
|
||||
@@ -1,7 +1,10 @@
|
||||
from pathlib import Path
|
||||
import sys
|
||||
import torch
|
||||
from train import CNN
|
||||
import numpy
|
||||
from PIL import Image, ImageFile
|
||||
import matplotlib.pyplot as plt
|
||||
from mnist import CNN
|
||||
|
||||
sys.path.append(str(Path(__file__).resolve().parent.parent.parent))
|
||||
import gpu_utils
|
||||
@@ -36,7 +39,7 @@ class Predictor:
|
||||
file_path = Path(__file__).resolve().parent.parent / 'models' / 'cnn.pth'
|
||||
self.cnn.load_state_dict(torch.load(file_path))
|
||||
|
||||
def predict(self, image: list[list[bool]]) -> PredictResult:
|
||||
def predict_sketchpad(self, image: list[list[bool]]) -> PredictResult:
|
||||
input = torch.Tensor(image).float().to(self.device)
|
||||
assert(input.dim() == 2)
|
||||
assert(input.size(0) == 28)
|
||||
@@ -51,4 +54,42 @@ class Predictor:
|
||||
with torch.no_grad():
|
||||
output = self.cnn(input)
|
||||
return PredictResult(output)
|
||||
|
||||
|
||||
def predict_image(self, image: ImageFile.ImageFile) -> PredictResult:
|
||||
# 确保图像为灰度图像，然后转换为numpy数组。
|
||||
# 注意这里的numpy数组是只读的，所以要先拷贝一份
|
||||
grayscale_image = image.convert('L')
|
||||
numpy_data = numpy.reshape(grayscale_image, (28, 28), copy=True)
|
||||
# 转换到Tensor，设置dtype并传到GPU上
|
||||
data = torch.from_numpy(numpy_data).float().to(self.device)
|
||||
# 归一化到255，又因为图像输入是白底黑字，需要做转换。
|
||||
data.div_(255.0).sub_(1).mul_(-1)
|
||||
|
||||
# 同理，挤出维度并预测
|
||||
input = data.unsqueeze(0).unsqueeze(0)
|
||||
with torch.no_grad():
|
||||
output = self.cnn(input)
|
||||
return PredictResult(output)
|
||||
|
||||
def main():
|
||||
predictor = Predictor()
|
||||
|
||||
# 遍历测试目录中的所有图片，并处理。
|
||||
test_dir = Path(__file__).resolve().parent.parent / 'test_images'
|
||||
for image_path in test_dir.glob('*.png'):
|
||||
if image_path.is_file():
|
||||
print(f'Predicting {image_path} ...')
|
||||
image = Image.open(image_path)
|
||||
rv = predictor.predict_image(image)
|
||||
|
||||
print(f'Predict digit: {rv.chosen_number()}')
|
||||
plt.figure(f'Image - {image_path}')
|
||||
plt.imshow(image)
|
||||
plt.axis('on')
|
||||
plt.title(f'Predict digit: {rv.chosen_number()}')
|
||||
plt.show()
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
main()
|
||||
|
||||
|
||||
@@ -169,7 +169,7 @@ class SketchpadApp:
|
||||
|
||||
def execute(self):
|
||||
"""执行按钮功能 - 将画板数据传递给后端"""
|
||||
prediction = self.predictor.predict(self.canvas_data)
|
||||
prediction = self.predictor.predict_sketchpad(self.canvas_data)
|
||||
self.show_in_table(prediction)
|
||||
|
||||
def reset(self):
|
||||
|
||||
@@ -1,145 +1,57 @@
|
||||
from pathlib import Path
|
||||
import sys
|
||||
import typing
|
||||
import numpy
|
||||
import torch
|
||||
from torch.utils.data import DataLoader, Dataset
|
||||
from torchvision.transforms import v2 as tvtrans
|
||||
import matplotlib.pyplot as plt
|
||||
import torch.nn.functional as F
|
||||
import torchinfo
|
||||
import ignite.engine
|
||||
import ignite.metrics
|
||||
from ignite.engine import Engine, Events
|
||||
from ignite.handlers.tqdm_logger import ProgressBar
|
||||
from mnist import CNN, MnistDataSource
|
||||
|
||||
sys.path.append(str(Path(__file__).resolve().parent.parent.parent))
|
||||
import gpu_utils
|
||||
|
||||
|
||||
class CNN(torch.nn.Module):
|
||||
"""卷积神经网络模型"""
|
||||
|
||||
def __init__(self):
|
||||
super(CNN, self).__init__()
|
||||
|
||||
self.conv1 = torch.nn.Conv2d(1, 32, kernel_size=(3, 3))
|
||||
self.pool1 = torch.nn.MaxPool2d(kernel_size=(2, 2))
|
||||
self.conv2 = torch.nn.Conv2d(32, 64, kernel_size=(3, 3))
|
||||
self.pool2 = torch.nn.MaxPool2d(kernel_size=(2, 2))
|
||||
self.conv3 = torch.nn.Conv2d(64, 64, kernel_size=(3, 3))
|
||||
self.flatten = torch.nn.Flatten()
|
||||
# 28x28过第一轮卷积后变为26x26，过第一轮池化后变为13x13
|
||||
# 过第二轮卷积后变为11x11，过第二轮池化后变为5x5
|
||||
# 过第三轮卷积后变为3x3。
|
||||
# 最后一轮卷积核个数为64。
|
||||
self.fc1 = torch.nn.Linear(64 * 3 * 3, 64)
|
||||
self.fc2 = torch.nn.Linear(64, 10)
|
||||
|
||||
def forward(self, x):
|
||||
x = F.relu(self.conv1(x))
|
||||
x = self.pool1(x)
|
||||
x = F.relu(self.conv2(x))
|
||||
x = self.pool2(x)
|
||||
x = F.relu(self.conv3(x))
|
||||
x = self.flatten(x)
|
||||
x = F.relu(self.fc1(x))
|
||||
x = self.fc2(x)
|
||||
return F.softmax(x, dim=1)
|
||||
|
||||
|
||||
class MnistDataset(Dataset):
|
||||
"""用于加载Mnist的自定义数据集"""
|
||||
|
||||
shape: int
|
||||
transform: tvtrans.Transform
|
||||
images_data: numpy.ndarray
|
||||
labels_data: torch.Tensor
|
||||
|
||||
def __init__(self, images: numpy.ndarray, labels: numpy.ndarray, transform: tvtrans.Transform):
|
||||
images_len: int = images.size(0)
|
||||
labels_len: int = labels.size(0)
|
||||
assert (images_len == labels_len)
|
||||
self.shape = images_len
|
||||
|
||||
self.images_data = images
|
||||
self.labels_data = torch.from_numpy(labels)
|
||||
self.transform = transform
|
||||
|
||||
def __getitem__(self, index):
|
||||
return self.transform(self.images_data[index]), self.labels_data[index]
|
||||
|
||||
def __len__(self):
|
||||
return self.shape
|
||||
|
||||
|
||||
class DataSource:
|
||||
"""用于读取MNIST数据的数据读取器"""
|
||||
|
||||
train_data: DataLoader
|
||||
test_data: DataLoader
|
||||
|
||||
def __init__(self, batch_size: int):
|
||||
datasets_path = Path(__file__).resolve().parent.parent / 'datasets' / 'mnist.npz'
|
||||
datasets = numpy.load(datasets_path)
|
||||
|
||||
# 所有图片均为黑底白字
|
||||
# 6万张训练图片：60000x28x28。标签只有第一维。
|
||||
train_images = datasets['x_train']
|
||||
train_labels = datasets['y_train']
|
||||
# 1万张测试图片：10000x28x28。标签只有第一维。
|
||||
test_images = datasets['x_test']
|
||||
test_labels = datasets['y_test']
|
||||
|
||||
# 定义数据转换器
|
||||
trans = tvtrans.Compose([
|
||||
# 从uint8转换为float32并自动归一化到0-1区间
|
||||
# tvtrans.ToTensor(),
|
||||
tvtrans.ToImage(),
|
||||
tvtrans.ToDtype(torch.float32, scale=True),
|
||||
# 为了符合后面图像的输入颜色通道条件，要在最后挤出一个新的维度
|
||||
#tvtrans.Lambda(lambda x: x.unsqueeze(-1))
|
||||
])
|
||||
|
||||
# 创建数据集
|
||||
train_dataset = MnistDataset(train_images,
|
||||
train_labels,
|
||||
transform=trans)
|
||||
test_dataset = MnistDataset(test_images, test_labels, transform=trans)
|
||||
|
||||
# 赋值到自身
|
||||
self.train_data = DataLoader(dataset=train_dataset,
|
||||
batch_size=batch_size,
|
||||
shuffle=False)
|
||||
self.test_data = DataLoader(dataset=test_dataset,
|
||||
batch_size=batch_size,
|
||||
shuffle=False)
|
||||
|
||||
|
||||
class Trainer:
|
||||
N_EPOCH: typing.ClassVar[int] = 5
|
||||
N_BATCH_SIZE: typing.ClassVar[int] = 1000
|
||||
|
||||
device: torch.device
|
||||
data_source: DataSource
|
||||
cnn: CNN
|
||||
data_source: MnistDataSource
|
||||
model: CNN
|
||||
|
||||
def __init__(self):
|
||||
self.device = gpu_utils.get_gpu_device()
|
||||
self.data_source = DataSource(Trainer.N_BATCH_SIZE)
|
||||
self.cnn = CNN().to(self.device)
|
||||
self.data_source = MnistDataSource(Trainer.N_BATCH_SIZE)
|
||||
self.model = CNN().to(self.device)
|
||||
# 展示模型结构。批次为指定批次数量，通道只有一个灰度通道，大小28x28。
|
||||
torchinfo.summary(self.model, (Trainer.N_BATCH_SIZE, 1, 28, 28))
|
||||
|
||||
def train(self):
|
||||
optimizer = torch.optim.Adam(self.cnn.parameters())
|
||||
optimizer = torch.optim.Adam(self.model.parameters(), eps=1e-7)
|
||||
# optimizer = torch.optim.AdamW(
|
||||
# self.model.parameters(),
|
||||
# lr=0.001, # 两者默认学习率都是 0.001
|
||||
# betas=(0.9, 0.999), # 两者默认值相同
|
||||
# eps=1e-07, # 【关键】匹配 TensorFlow 的默认 epsilon
|
||||
# weight_decay=0.0, # 两者默认都是 0
|
||||
# amsgrad=False # 两者默认都是 False
|
||||
# )
|
||||
loss_func = torch.nn.CrossEntropyLoss()
|
||||
|
||||
for epoch in range(Trainer.N_EPOCH):
|
||||
self.cnn.train()
|
||||
self.model.train()
|
||||
|
||||
batch_images: torch.Tensor
|
||||
batch_labels: torch.Tensor
|
||||
for batch_index, (batch_images, batch_labels) in enumerate(self.data_source.train_data):
|
||||
for batch_index, (batch_images, batch_labels) in enumerate(self.data_source.train_loader):
|
||||
gpu_images = batch_images.to(self.device)
|
||||
gpu_labels = batch_labels.to(self.device)
|
||||
|
||||
optimizer.zero_grad()
|
||||
prediction: torch.Tensor = self.cnn(gpu_images)
|
||||
prediction: torch.Tensor = self.model(gpu_images)
|
||||
loss: torch.Tensor = loss_func(prediction, gpu_labels)
|
||||
optimizer.zero_grad()
|
||||
loss.backward()
|
||||
optimizer.step()
|
||||
|
||||
@@ -151,20 +63,20 @@ class Trainer:
|
||||
file_dir_path = Path(__file__).resolve().parent.parent / 'models'
|
||||
file_dir_path.mkdir(parents=True, exist_ok=True)
|
||||
file_path = file_dir_path / 'cnn.pth'
|
||||
torch.save(self.cnn.state_dict(), file_path)
|
||||
torch.save(self.model.state_dict(), file_path)
|
||||
print(f'模型已保存至：{file_path}')
|
||||
|
||||
def test(self):
|
||||
self.cnn.eval()
|
||||
self.model.eval()
|
||||
correct_sum = 0
|
||||
total_sum = 0
|
||||
|
||||
with torch.no_grad():
|
||||
for batch_images, batch_labels in self.data_source.test_data:
|
||||
for batch_images, batch_labels in self.data_source.test_loader:
|
||||
gpu_images = batch_images.to(self.device)
|
||||
gpu_labels = batch_labels.to(self.device)
|
||||
|
||||
possibilities: torch.Tensor = self.cnn(gpu_images)
|
||||
possibilities: torch.Tensor = self.model(gpu_images)
|
||||
# 输出出来是10个数字各自的可能性，所以要选取最高可能性的那个对比
|
||||
# 在dim=1上找最大的那个，就选那个。dim=0是批次所以不管他。
|
||||
_, prediction = possibilities.max(1)
|
||||
@@ -175,13 +87,105 @@ class Trainer:
|
||||
test_acc = 100. * correct_sum / total_sum
|
||||
print(f"准确率: {test_acc:.4f}%，共测试了{total_sum}张图片")
|
||||
|
||||
|
||||
def main():
|
||||
trainer = Trainer()
|
||||
trainer.train()
|
||||
trainer.save()
|
||||
trainer.test()
|
||||
|
||||
# N_EPOCH: int = 5
|
||||
# N_BATCH_SIZE: int = 1000
|
||||
# N_LOG_INTERVAL: int = 10
|
||||
|
||||
# class Trainer:
|
||||
|
||||
# device: torch.device
|
||||
# data_source: MnistDataSource
|
||||
# model: CNN
|
||||
|
||||
# trainer: Engine
|
||||
# train_evaluator: Engine
|
||||
# test_evaluator: Engine
|
||||
|
||||
# def __init__(self):
|
||||
# self.device = gpu_utils.get_gpu_device()
|
||||
# self.model = CNN().to(self.device)
|
||||
# self.data_source = MnistDataSource(batch_size=N_BATCH_SIZE)
|
||||
# # 展示模型结构。批次为指定批次数量，通道只有一个灰度通道，大小28x28。
|
||||
# torchinfo.summary(self.model, (N_BATCH_SIZE, 1, 28, 28))
|
||||
|
||||
# #optimizer = torch.optim.Adam(self.model.parameters(), eps=1e-7)
|
||||
# optimizer = torch.optim.AdamW(
|
||||
# self.model.parameters(),
|
||||
# lr=0.001, # 两者默认学习率都是 0.001
|
||||
# betas=(0.9, 0.999), # 两者默认值相同
|
||||
# eps=1e-07, # 【关键】匹配 TensorFlow 的默认 epsilon
|
||||
# weight_decay=0.0, # 两者默认都是 0
|
||||
# amsgrad=False # 两者默认都是 False
|
||||
# )
|
||||
# criterion = torch.nn.CrossEntropyLoss()
|
||||
# self.trainer = ignite.engine.create_supervised_trainer(
|
||||
# self.model, optimizer, criterion, self.device
|
||||
# )
|
||||
|
||||
# eval_metrics = {
|
||||
# "accuracy": ignite.metrics.Accuracy(device=self.device),
|
||||
# "loss": ignite.metrics.Loss(criterion, device=self.device)
|
||||
# }
|
||||
# self.train_evaluator = ignite.engine.create_supervised_evaluator(
|
||||
# self.model, metrics=eval_metrics, device=self.device)
|
||||
# self.test_evaluator = ignite.engine.create_supervised_evaluator(
|
||||
# self.model, metrics=eval_metrics, device=self.device)
|
||||
|
||||
# self.trainer.add_event_handler(
|
||||
# Events.ITERATION_COMPLETED(every=N_LOG_INTERVAL),
|
||||
# lambda engine: self.log_intrain_loss(engine)
|
||||
# )
|
||||
# self.trainer.add_event_handler(
|
||||
# Events.EPOCH_COMPLETED,
|
||||
# lambda trainer: self.log_train_results(trainer)
|
||||
# )
|
||||
# self.trainer.add_event_handler(
|
||||
# Events.COMPLETED,
|
||||
# lambda _: self.log_test_results()
|
||||
# )
|
||||
# self.trainer.add_event_handler(
|
||||
# Events.COMPLETED,
|
||||
# lambda _: self.save_model()
|
||||
# )
|
||||
|
||||
# progressbar = ProgressBar()
|
||||
# progressbar.attach(self.trainer)
|
||||
|
||||
# def run(self):
|
||||
# self.trainer.run(self.data_source.train_loader, max_epochs=N_EPOCH)
|
||||
|
||||
# def log_intrain_loss(self, engine: Engine):
|
||||
# print(f"Epoch: {engine.state.epoch}, Loss: {engine.state.output:.4f}\r", end="")
|
||||
|
||||
# def log_train_results(self, trainer: Engine):
|
||||
# self.train_evaluator.run(self.data_source.train_loader)
|
||||
# metrics = self.train_evaluator.state.metrics
|
||||
# print()
|
||||
# print(f"Training - Epoch: {trainer.state.epoch}, Avg Accuracy: {metrics['accuracy']:.4f}, Avg Loss: {metrics['loss']:.4f}")
|
||||
|
||||
# def log_test_results(self):
|
||||
# self.test_evaluator.run(self.data_source.test_loader)
|
||||
# metrics = self.test_evaluator.state.metrics
|
||||
# print(f"Test - Avg Accuracy: {metrics['accuracy']:.4f} Avg Loss: {metrics['loss']:.4f}")
|
||||
|
||||
# def save_model(self):
|
||||
# file_dir_path = Path(__file__).resolve().parent.parent / 'models'
|
||||
# file_dir_path.mkdir(parents=True, exist_ok=True)
|
||||
# file_path = file_dir_path / 'cnn.pth'
|
||||
# torch.save(self.model.state_dict(), file_path)
|
||||
# print(f'Model was saved into: {file_path}')
|
||||
|
||||
# def main():
|
||||
# trainer = Trainer()
|
||||
# trainer.run()
|
||||
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
gpu_utils.print_gpu_availability()
|
||||
|
||||
2
exp2/test_images/.gitignore
vendored
Normal file
2
exp2/test_images/.gitignore
vendored
Normal file
@@ -0,0 +1,2 @@
|
||||
# Ignore all test images
|
||||
*.png
|
||||
Reference in New Issue
Block a user