第四次作业：CNN实战
第四次作业：CNN实战
0x00样例
model_vgg = models.vgg16(pretrained=True)
with open('./imagenet_class_index.json') as f:
class_dict = json.load(f)
dic_imagenet = [class_dict[str(i)][1] for i in range(len(class_dict))]
inputs_try , labels_try = inputs_try.to(device), labels_try.to(device)
model_vgg = model_vgg.to(device)
outputs_try = model_vgg(inputs_try)
print(outputs_try)
print(outputs_try.shape)
'''
可以看到结果为5行，1000列的数据，每一列代表对每一种目标识别的结果。
但是我也可以观察到，结果非常奇葩，有负数，有正数，
为了将VGG网络输出的结果转化为对每一类的预测概率，我们把结果输入到 Softmax 函数
'''
m_softm = nn.Softmax(dim=1)
probs = m_softm(outputs_try)
vals_try,pred_try = torch.max(probs,dim=1)
print( 'prob sum: ', torch.sum(probs,1))
print( 'vals_try: ', vals_try)
print( 'pred_try: ', pred_try)
print([dic_imagenet[i] for i in pred_try.data])
imshow(torchvision.utils.make_grid(inputs_try.data.cpu()),
title=[dset_classes[x] for x in labels_try.data.cpu()])
下载torchvision中集成的vgg在ImageNet上的预训练模型，并且下载了ImageNet的1000分类的标签。将valid中第一个batch的图片放入预训练模型中进行测试。同时用softmax对预测概率进行归一化。得到预测结果
对五张图片预测得到的结果分别是仓鼠、斑猫、斑猫、斑猫、黑足雪貂。
print(model_vgg)
model_vgg_new = model_vgg;
for param in model_vgg_new.parameters():
param.requires_grad = False
model_vgg_new.classifier._modules['6'] = nn.Linear(4096, 2)
model_vgg_new.classifier._modules['7'] = torch.nn.LogSoftmax(dim = 1)
model_vgg_new = model_vgg_new.to(device)
print(model_vgg_new.classifier)
修改预训练的vgg的网络结构。首先param.requires_grad = Flases来冻结预训练模型中的权重，只对最后一个全连接层进行训练。
全连接层的out_features从1000改为2，以便于适应当前的二分类任务。同时加了一层Softmax函数来归一化。
'''
第一步：创建损失函数和优化器
损失函数 NLLLoss() 的 输入 是一个对数概率向量和一个目标标签.
它不会为我们计算对数概率，适合最后一层是log_softmax()的网络.
'''
criterion = nn.NLLLoss()
# 学习率
lr = 0.001
# 随机梯度下降
optimizer_vgg = torch.optim.SGD(model_vgg_new.classifier[6].parameters(),lr = lr)
'''
第二步：训练模型
'''
def train_model(model,dataloader,size,epochs=1,optimizer=None):
model.train()
for epoch in range(epochs):
running_loss = 0.0
running_corrects = 0
count = 0
for inputs,classes in dataloader:
inputs = inputs.to(device)
classes = classes.to(device)
outputs = model(inputs)
loss = criterion(outputs,classes)
optimizer = optimizer
optimizer.zero_grad()
loss.backward()
optimizer.step()
_,preds = torch.max(outputs.data,1)
# statistics
running_loss += loss.data.item()
running_corrects += torch.sum(preds == classes.data)
count += len(inputs)
if count % 256==0:
print('Epoch: ', epoch, 'Training: No. ', count, ' process ... total: ', size)
epoch_loss = running_loss / size
epoch_acc = running_corrects.data.item() / size
print('Loss: {:.4f} Acc: {:.4f}'.format(
epoch_loss, epoch_acc))
# 模型训练
train_model(model_vgg_new,loader_train,size=dset_sizes['train'], epochs=1,
optimizer=optimizer_vgg)
对冻结后的模型进行训练。训练一个epoch
def test_model(model,dataloader,size):
model.eval()
predictions = np.zeros(size)
all_classes = np.zeros(size)
all_proba = np.zeros((size,2))
i = 0
running_loss = 0.0
running_corrects = 0
for inputs,classes in dataloader:
inputs = inputs.to(device)
classes = classes.to(device)
outputs = model(inputs)
loss = criterion(outputs,classes)
_,preds = torch.max(outputs.data,1)
# statistics
running_loss += loss.data.item()
running_corrects += torch.sum(preds == classes.data)
predictions[i:i+len(classes)] = preds.to('cpu').numpy()
all_classes[i:i+len(classes)] = classes.to('cpu').numpy()
all_proba[i:i+len(classes),:] = outputs.data.to('cpu').numpy()
i += len(classes)
print('Testing: No. ', i, ' process ... total: ', size)
epoch_loss = running_loss / size
epoch_acc = running_corrects.data.item() / size
print('Loss: {:.4f} Acc: {:.4f}'.format(
epoch_loss, epoch_acc))
return predictions, all_proba, all_classes
predictions, all_proba, all_classes = test_model(model_vgg_new,loader_valid,size=dset_sizes['valid'])
测试得到的准确率为95.80%
# 单次可视化显示的图片个数
n_view = 8
wrongs = np.where(predictions!=all_classes)[0]
from numpy.random import random, permutation
idx = permutation(wrongs)[:n_view]
print('random wrong idx: ', idx)
loader_wrong = torch.utils.data.DataLoader([dsets['valid'][x] for x in idx],
batch_size = n_view,shuffle=True)
for data in loader_correct:
inputs_wrg,labels_wrg = data
# Make a grid from batch
out = torchvision.utils.make_grid(inputs_wrg)
imshow(out, title=[l.item() for l in labels_wrg])
随机查看一些预测错误的数据
0x01 测试
import numpy as np
import os
import glob
import torch
import torch.nn as nn
import torchvision
from torchvision import models, transforms, datasets
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
print('Using gpu: %s ' % torch.cuda.is_available())
normalize = transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
vgg_format = transforms.Compose([
transforms.CenterCrop(224),
transforms.ToTensor(),
normalize,
])
data_dir = './data/cat_dog'
dsets = {x: datasets.ImageFolder(os.path.join(data_dir, x), vgg_format)
for x in ['train', 'val']}
dset_sizes = {x: len(dsets[x]) for x in ['train', 'val']}
dset_classes = dsets['train'].classes
loader_train = torch.utils.data.DataLoader(dsets['train'], batch_size=128, shuffle=True, num_workers=2)
loader_valid = torch.utils.data.DataLoader(dsets['val'], batch_size=32, shuffle=False, num_workers=2)
model_vgg = models.vgg16(pretrained=True)
print(model_vgg)
model_vgg_new = model_vgg;
for param in model_vgg_new.parameters():
param.requires_grad = False
model_vgg_new.classifier._modules['6'] = nn.Linear(4096, 2)
model_vgg_new.classifier._modules['7'] = torch.nn.LogSoftmax(dim = 1)
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
print('Using gpu: %s ' % torch.cuda.is_available())
model_vgg_new = model_vgg_new.to(device)
print(model_vgg_new.classifier)
criterion = nn.NLLLoss()
# 学习率
lr = 0.001
# 随机梯度下降
optimizer_vgg = torch.optim.Adam(model_vgg_new.classifier.parameters(),lr = lr)
def adjust_learning_rate(optimizer, epoch):
global lr
l_r = lr * (0.1 ** (epoch // 30))
for param_group in optimizer.param_groups:
print(param_group)
param_group['lr'] = l_r
def train_model(model,dataloader,size,epochs=80,optimizer=None):
model.train()
for epoch in range(epochs):
adjust_learning_rate(optimizer_vgg, epoch)
running_loss = 0.0
running_corrects = 0
count = 0
for inputs,classes in dataloader:
inputs = inputs.to(device)
classes = classes.to(device)
outputs = model(inputs)
loss = criterion(outputs,classes)
optimizer = optimizer
optimizer.zero_grad()
loss.backward()
optimizer.step()
_,preds = torch.max(outputs.data,1)
# statistics
running_loss += loss.data.item()
running_corrects += torch.sum(preds == classes.data)
count += len(inputs)
if count % 2560==0:
print('Epoch: ', epoch, 'LR:', optimizer.param_groups[0]['lr'], 'Training: No. ', count, ' process ... total: ', size)
epoch_loss = running_loss / size
epoch_acc = running_corrects.data.item() / size
print('Loss: {:.4f} Acc: {:.4f}'.format(
epoch_loss*10000, epoch_acc))
if epoch % 10 == 0:
torch.save(model, "./vgg_ep" + str(epoch) + ".pth")
# 模型训练
train_model(model_vgg_new,loader_train,size=dset_sizes['train'], epochs=301,
optimizer=optimizer_vgg)
def adjust_learning_rate(optimizer, epoch):
"""
动态调整学习率
"""
global lr
l_r = lr * (0.1 ** (epoch // 30))
for param_group in optimizer.param_groups:
param_group['lr'] = l_r
-
将优化器从SGD改为Adam
-
采用动态学习率，每训练30个epoch将学习率调小（*0.1）
-
将epochs调整到301
效果98.45%
浙公网安备 33010602011771号