- 🍨 本文为🔗365天深度学习训练营 中的学习记录博客
- 🍖 原作者：K同学啊
代码
import matplotlib.pyplot as plt
#隐藏警告
import warnings
warnings.filterwarnings("ignore") #忽略警告信息
plt.rcParams['font.sans-serif'] = ['SimHei'] # 用来正常显示中文标签
plt.rcParams['axes.unicode_minus'] = False # 用来正常显示负号
plt.rcParams['figure.dpi'] = 100 #分辨率
epochs_range = range(epochs)
plt.figure(figsize=(12, 3))
plt.subplot(1, 2, 1)
plt.plot(epochs_range, train_acc, label='Training Accuracy')
plt.plot(epochs_range, test_acc, label='Test Accuracy')
plt.legend(loc='lower right')
plt.title('Training and Validation Accuracy')
plt.subplot(1, 2, 2)
plt.plot(epochs_range, train_loss, label='Training Loss')
plt.plot(epochs_range, test_loss, label='Test Loss')
plt.legend(loc='upper right')
plt.title('Training and Validation Loss')
plt.show()
from PIL import Image
classes = list(total_data.class_to_idx)
def predict_one_image(image_path, model, transform, classes):
test_img = Image.open(image_path).convert('RGB')
plt.imshow(test_img) # 展示预测的图片
test_img = transform(test_img)
img = test_img.to(device).unsqueeze(0)
model.eval()
output = model(img)
_,pred = torch.max(output,1)
pred_class = classes[pred]
print(f'预测结果是：{pred_class}')
# 预测训练集中的某张照片
predict_one_image(image_path='./data/p/p7-data/Dark/dark (99).png',
model=model,
transform=train_transforms,
classes=classes)
预测结果是：Dark
训练结果
VGG16训练结果：
调整为SGD
调整网络结构，略微减少了一些参数
==========================================================================================
Layer (type:depth-idx) Output Shape Param #
==========================================================================================
├─Sequential: 1-1 [-1, 64, 112, 112] --
| └─Conv2d: 2-1 [-1, 64, 224, 224] 1,792
| └─ReLU: 2-2 [-1, 64, 224, 224] --
| └─Conv2d: 2-3 [-1, 64, 224, 224] 36,928
| └─ReLU: 2-4 [-1, 64, 224, 224] --
| └─MaxPool2d: 2-5 [-1, 64, 112, 112] --
├─Sequential: 1-2 [-1, 128, 56, 56] --
| └─Conv2d: 2-6 [-1, 128, 112, 112] 73,856
| └─ReLU: 2-7 [-1, 128, 112, 112] --
| └─Conv2d: 2-8 [-1, 128, 112, 112] 147,584
| └─ReLU: 2-9 [-1, 128, 112, 112] --
| └─MaxPool2d: 2-10 [-1, 128, 56, 56] --
├─Sequential: 1-3 [-1, 256, 28, 28] --
| └─Conv2d: 2-11 [-1, 256, 56, 56] 295,168
| └─ReLU: 2-12 [-1, 256, 56, 56] --
| └─Conv2d: 2-13 [-1, 256, 56, 56] 590,080
| └─ReLU: 2-14 [-1, 256, 56, 56] --
| └─MaxPool2d: 2-15 [-1, 256, 28, 28] --
├─Sequential: 1-4 [-1, 512, 14, 14] --
| └─Conv2d: 2-16 [-1, 512, 28, 28] 1,180,160
| └─ReLU: 2-17 [-1, 512, 28, 28] --
| └─Conv2d: 2-18 [-1, 512, 28, 28] 2,359,808
| └─ReLU: 2-19 [-1, 512, 28, 28] --
| └─MaxPool2d: 2-20 [-1, 512, 14, 14] --
├─Sequential: 1-5 [-1, 512, 7, 7] --
| └─Conv2d: 2-21 [-1, 512, 14, 14] 2,359,808
| └─ReLU: 2-22 [-1, 512, 14, 14] --
| └─Conv2d: 2-23 [-1, 512, 14, 14] 2,359,808
| └─ReLU: 2-24 [-1, 512, 14, 14] --
| └─MaxPool2d: 2-25 [-1, 512, 7, 7] --
├─Sequential: 1-6 [-1, 4] --
| └─Linear: 2-26 [-1, 4096] 102,764,544
| └─ReLU: 2-27 [-1, 4096] --
| └─Linear: 2-28 [-1, 4096] 16,781,312
| └─ReLU: 2-29 [-1, 4096] --
| └─Linear: 2-30 [-1, 4] 16,388
==========================================================================================
Total params: 128,967,236
Trainable params: 128,967,236
Non-trainable params: 0
Total mult-adds (G): 11.43
==========================================================================================
Input size (MB): 0.57
Forward/backward pass size (MB): 93.47
Params size (MB): 491.97
Estimated Total Size (MB): 586.01
==========================================================================================
从训练结果上来说，基本和满血VGG16一致
降参数
调小卷积核，增加卷积层，减小全连接层，能一定程度减少参数
Using cuda device
==========================================================================================
Layer (type:depth-idx) Output Shape Param #
==========================================================================================
├─Sequential: 1-1 [-1, 64, 112, 112] --
| └─Conv2d: 2-1 [-1, 64, 224, 224] 1,792
| └─ReLU: 2-2 [-1, 64, 224, 224] --
| └─Conv2d: 2-3 [-1, 64, 224, 224] 36,928
| └─ReLU: 2-4 [-1, 64, 224, 224] --
| └─MaxPool2d: 2-5 [-1, 64, 112, 112] --
├─Sequential: 1-2 [-1, 128, 56, 56] --
| └─Conv2d: 2-6 [-1, 128, 112, 112] 73,856
| └─ReLU: 2-7 [-1, 128, 112, 112] --
| └─MaxPool2d: 2-8 [-1, 128, 56, 56] --
├─Sequential: 1-3 [-1, 256, 28, 28] --
| └─Conv2d: 2-9 [-1, 256, 56, 56] 295,168
| └─ReLU: 2-10 [-1, 256, 56, 56] --
| └─Conv2d: 2-11 [-1, 256, 56, 56] 590,080
| └─ReLU: 2-12 [-1, 256, 56, 56] --
| └─MaxPool2d: 2-13 [-1, 256, 28, 28] --
├─Sequential: 1-4 [-1, 512, 15, 15] --
| └─Conv2d: 2-14 [-1, 512, 29, 29] 524,800
| └─ReLU: 2-15 [-1, 512, 29, 29] --
| └─Conv2d: 2-16 [-1, 512, 30, 30] 1,049,088
| └─ReLU: 2-17 [-1, 512, 30, 30] --
| └─Conv2d: 2-18 [-1, 512, 31, 31] 1,049,088
| └─ReLU: 2-19 [-1, 512, 31, 31] --
| └─MaxPool2d: 2-20 [-1, 512, 15, 15] --
├─Sequential: 1-5 [-1, 512, 9, 9] --
| └─Conv2d: 2-21 [-1, 512, 16, 16] 1,049,088
| └─ReLU: 2-22 [-1, 512, 16, 16] --
| └─Conv2d: 2-23 [-1, 512, 17, 17] 1,049,088
| └─ReLU: 2-24 [-1, 512, 17, 17] --
| └─Conv2d: 2-25 [-1, 512, 18, 18] 1,049,088
| └─ReLU: 2-26 [-1, 512, 18, 18] --
| └─MaxPool2d: 2-27 [-1, 512, 9, 9] --
├─Sequential: 1-6 [-1, 512, 3, 3] --
| └─Conv2d: 2-28 [-1, 512, 9, 9] 2,359,808
| └─ReLU: 2-29 [-1, 512, 9, 9] --
| └─MaxPool2d: 2-30 [-1, 512, 3, 3] --
├─Sequential: 1-7 [-1, 4] --
| └─Linear: 2-31 [-1, 4096] 18,878,464
| └─ReLU: 2-32 [-1, 4096] --
| └─Linear: 2-33 [-1, 4096] 16,781,312
| └─ReLU: 2-34 [-1, 4096] --
| └─Linear: 2-35 [-1, 4] 16,388
==========================================================================================
Total params: 44,804,036
Trainable params: 44,804,036
Non-trainable params: 0
Total mult-adds (G): 9.21
==========================================================================================
Input size (MB): 0.57
Forward/backward pass size (MB): 87.83
Params size (MB): 170.91
Estimated Total Size (MB): 259.32
==========================================================================================
==========================================================================================
Layer (type:depth-idx) Output Shape Param #
==========================================================================================
├─Sequential: 1-1 [-1, 64, 112, 112] --
| └─Conv2d: 2-1 [-1, 64, 224, 224] 1,792
| └─ReLU: 2-2 [-1, 64, 224, 224] --
| └─Conv2d: 2-3 [-1, 64, 224, 224] 36,928
| └─ReLU: 2-4 [-1, 64, 224, 224] --
| └─MaxPool2d: 2-5 [-1, 64, 112, 112] --
├─Sequential: 1-2 [-1, 128, 56, 56] --
| └─Conv2d: 2-6 [-1, 128, 112, 112] 73,856
| └─ReLU: 2-7 [-1, 128, 112, 112] --
| └─MaxPool2d: 2-8 [-1, 128, 56, 56] --
├─Sequential: 1-3 [-1, 256, 28, 28] --
| └─Conv2d: 2-9 [-1, 256, 56, 56] 295,168
| └─ReLU: 2-10 [-1, 256, 56, 56] --
| └─Conv2d: 2-11 [-1, 256, 56, 56] 590,080
| └─ReLU: 2-12 [-1, 256, 56, 56] --
| └─MaxPool2d: 2-13 [-1, 256, 28, 28] --
├─Sequential: 1-4 [-1, 512, 15, 15] --
| └─Conv2d: 2-14 [-1, 512, 29, 29] 524,800
| └─ReLU: 2-15 [-1, 512, 29, 29] --
| └─Conv2d: 2-16 [-1, 512, 30, 30] 1,049,088
| └─ReLU: 2-17 [-1, 512, 30, 30] --
| └─Conv2d: 2-18 [-1, 512, 31, 31] 1,049,088
| └─ReLU: 2-19 [-1, 512, 31, 31] --
| └─MaxPool2d: 2-20 [-1, 512, 15, 15] --
├─Sequential: 1-5 [-1, 512, 9, 9] --
| └─Conv2d: 2-21 [-1, 512, 16, 16] 1,049,088
| └─ReLU: 2-22 [-1, 512, 16, 16] --
| └─Conv2d: 2-23 [-1, 512, 17, 17] 1,049,088
| └─ReLU: 2-24 [-1, 512, 17, 17] --
| └─Conv2d: 2-25 [-1, 512, 18, 18] 1,049,088
| └─ReLU: 2-26 [-1, 512, 18, 18] --
| └─MaxPool2d: 2-27 [-1, 512, 9, 9] --
├─Sequential: 1-6 [-1, 512, 3, 3] --
| └─Conv2d: 2-28 [-1, 512, 9, 9] 2,359,808
| └─ReLU: 2-29 [-1, 512, 9, 9] --
| └─MaxPool2d: 2-30 [-1, 512, 3, 3] --
├─Sequential: 1-7 [-1, 4] --
| └─Linear: 2-31 [-1, 4096] 18,878,464
| └─ReLU: 2-32 [-1, 4096] --
| └─Linear: 2-33 [-1, 4096] 16,781,312
| └─ReLU: 2-34 [-1, 4096] --
| └─Linear: 2-35 [-1, 4] 16,388
==========================================================================================
Total params: 44,804,036
Trainable params: 44,804,036
Non-trainable params: 0
Total mult-adds (G): 9.21
==========================================================================================
Input size (MB): 0.57
Forward/backward pass size (MB): 87.83
Params size (MB): 170.91
Estimated Total Size (MB): 259.32
==========================================================================================
参数情况：从134,276,932降到了44,804,036，数量是原来的1/3，测试结果尚可
Epoch: 1, Train_acc:24.4%, Train_loss:1.389, Test_acc:23.8%, Test_loss:1.386, Lr:1.00E-04
Epoch: 2, Train_acc:39.4%, Train_loss:1.173, Test_acc:47.9%, Test_loss:1.004, Lr:1.00E-04
Epoch: 3, Train_acc:56.5%, Train_loss:0.768, Test_acc:59.2%, Test_loss:0.831, Lr:1.00E-04
Epoch: 4, Train_acc:72.3%, Train_loss:0.604, Test_acc:81.2%, Test_loss:0.548, Lr:1.00E-04
Epoch: 5, Train_acc:85.5%, Train_loss:0.334, Test_acc:69.6%, Test_loss:0.862, Lr:1.00E-04
Epoch: 6, Train_acc:82.2%, Train_loss:0.407, Test_acc:83.8%, Test_loss:0.400, Lr:1.00E-04
Epoch: 7, Train_acc:91.6%, Train_loss:0.229, Test_acc:90.0%, Test_loss:0.253, Lr:1.00E-04
Epoch: 8, Train_acc:90.6%, Train_loss:0.240, Test_acc:85.4%, Test_loss:0.366, Lr:1.00E-04
Epoch: 9, Train_acc:94.2%, Train_loss:0.164, Test_acc:93.8%, Test_loss:0.186, Lr:1.00E-04
Epoch:10, Train_acc:94.1%, Train_loss:0.146, Test_acc:93.3%, Test_loss:0.161, Lr:1.00E-04
Epoch:11, Train_acc:96.8%, Train_loss:0.082, Test_acc:94.6%, Test_loss:0.138, Lr:1.00E-04
Epoch:12, Train_acc:97.5%, Train_loss:0.073, Test_acc:94.2%, Test_loss:0.203, Lr:1.00E-04
Epoch:13, Train_acc:98.1%, Train_loss:0.058, Test_acc:94.6%, Test_loss:0.229, Lr:1.00E-04
Epoch:14, Train_acc:95.3%, Train_loss:0.133, Test_acc:95.4%, Test_loss:0.150, Lr:1.00E-04
Epoch:15, Train_acc:98.9%, Train_loss:0.047, Test_acc:96.7%, Test_loss:0.133, Lr:1.00E-04
Epoch:16, Train_acc:94.2%, Train_loss:0.153, Test_acc:90.8%, Test_loss:0.187, Lr:1.00E-04
Epoch:17, Train_acc:97.0%, Train_loss:0.074, Test_acc:97.9%, Test_loss:0.063, Lr:1.00E-04
Epoch:18, Train_acc:97.5%, Train_loss:0.059, Test_acc:97.1%, Test_loss:0.071, Lr:1.00E-04
Epoch:19, Train_acc:98.9%, Train_loss:0.029, Test_acc:96.7%, Test_loss:0.105, Lr:1.00E-04
Epoch:20, Train_acc:97.9%, Train_loss:0.051, Test_acc:97.1%, Test_loss:0.068, Lr:1.00E-04
Epoch:21, Train_acc:98.2%, Train_loss:0.043, Test_acc:97.9%, Test_loss:0.061, Lr:1.00E-04
Epoch:22, Train_acc:98.2%, Train_loss:0.042, Test_acc:96.7%, Test_loss:0.131, Lr:1.00E-04
Epoch:23, Train_acc:99.3%, Train_loss:0.029, Test_acc:98.3%, Test_loss:0.049, Lr:1.00E-04
Epoch:24, Train_acc:99.7%, Train_loss:0.011, Test_acc:97.9%, Test_loss:0.084, Lr:1.00E-04
Epoch:25, Train_acc:99.7%, Train_loss:0.008, Test_acc:97.9%, Test_loss:0.048, Lr:1.00E-04
Epoch:26, Train_acc:95.9%, Train_loss:0.147, Test_acc:95.8%, Test_loss:0.127, Lr:1.00E-04
Epoch:27, Train_acc:99.0%, Train_loss:0.044, Test_acc:96.7%, Test_loss:0.057, Lr:1.00E-04
Epoch:28, Train_acc:95.0%, Train_loss:0.161, Test_acc:91.7%, Test_loss:0.226, Lr:1.00E-04
Epoch:29, Train_acc:98.5%, Train_loss:0.049, Test_acc:96.2%, Test_loss:0.100, Lr:1.00E-04
Epoch:30, Train_acc:99.4%, Train_loss:0.024, Test_acc:97.5%, Test_loss:0.054, Lr:1.00E-04
Epoch:31, Train_acc:99.2%, Train_loss:0.018, Test_acc:96.2%, Test_loss:0.136, Lr:1.00E-04
Epoch:32, Train_acc:97.8%, Train_loss:0.060, Test_acc:96.2%, Test_loss:0.104, Lr:1.00E-04
Epoch:33, Train_acc:97.3%, Train_loss:0.078, Test_acc:94.6%, Test_loss:0.157, Lr:1.00E-04
Epoch:34, Train_acc:98.1%, Train_loss:0.059, Test_acc:95.8%, Test_loss:0.350, Lr:1.00E-04
Epoch:35, Train_acc:98.3%, Train_loss:0.038, Test_acc:97.5%, Test_loss:0.128, Lr:1.00E-04
Epoch:36, Train_acc:99.1%, Train_loss:0.030, Test_acc:97.9%, Test_loss:0.061, Lr:1.00E-04
Epoch:37, Train_acc:98.6%, Train_loss:0.039, Test_acc:96.2%, Test_loss:0.116, Lr:1.00E-04
Epoch:38, Train_acc:98.6%, Train_loss:0.040, Test_acc:96.2%, Test_loss:0.136, Lr:1.00E-04
Epoch:39, Train_acc:99.9%, Train_loss:0.005, Test_acc:97.9%, Test_loss:0.070, Lr:1.00E-04
Epoch:40, Train_acc:97.8%, Train_loss:0.082, Test_acc:96.7%, Test_loss:0.129, Lr:1.00E-04
Done
扫一扫
1万+
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
