{ "cells": [ { "cell_type": "markdown", "metadata": {}, "source": [ "# 9.1 图像增广" ] }, { "cell_type": "code", "execution_count": 1, "metadata": { "ExecuteTime": { "end_time": "2019-05-29T14:27:44.323427Z", "start_time": "2019-05-29T14:27:43.032842Z" } }, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "1.0.0\n", "cuda\n" ] } ], "source": [ "%matplotlib inline\n", "import os\n", "import time\n", "import torch\n", "from torch import nn, optim\n", "from torch.utils.data import Dataset, DataLoader\n", "import torchvision\n", "from PIL import Image\n", "\n", "os.environ[\"CUDA_VISIBLE_DEVICES\"] = \"0\"\n", "\n", "import sys\n", "sys.path.append(\"..\") \n", "import d2lzh_pytorch as d2l\n", "device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')\n", "print(torch.__version__)\n", "print(device)" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## 9.1.1 常用的图像增广方法" ] }, { "cell_type": "code", "execution_count": 2, "metadata": { "ExecuteTime": { "end_time": "2019-05-29T14:27:44.619786Z", "start_time": "2019-05-29T14:27:44.326107Z" } }, "outputs": [ { "data": { "text/plain": [ "" ] }, "execution_count": 2, "metadata": {}, "output_type": "execute_result" }, { "data": { "image/svg+xml": [ "\n", "\n", "\n", "\n" ], "text/plain": [ "

" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" } ], "source": [ "d2l.set_figsize()\n", "img = Image.open('../../img/cat1.jpg')\n", "d2l.plt.imshow(img)" ] }, { "cell_type": "code", "execution_count": 3, "metadata": { "ExecuteTime": { "end_time": "2019-05-29T14:27:44.625705Z", "start_time": "2019-05-29T14:27:44.621800Z" } }, "outputs": [], "source": [ "# 本函数已保存在d2lzh_pytorch包中方便以后使用\n", "def show_images(imgs, num_rows, num_cols, scale=2):\n", " figsize = (num_cols * scale, num_rows * scale)\n", " _, axes = d2l.plt.subplots(num_rows, num_cols, figsize=figsize)\n", " for i in range(num_rows):\n", " for j in range(num_cols):\n", " axes[i][j].imshow(imgs[i * num_cols + j])\n", " axes[i][j].axes.get_xaxis().set_visible(False)\n", " axes[i][j].axes.get_yaxis().set_visible(False)\n", " return axes" ] }, { "cell_type": "code", "execution_count": 4, "metadata": { "ExecuteTime": { "end_time": "2019-05-29T14:27:44.633413Z", "start_time": "2019-05-29T14:27:44.627482Z" } }, "outputs": [], "source": [ "def apply(img, aug, num_rows=2, num_cols=4, scale=1.5):\n", " Y = [aug(img) for _ in range(num_rows * num_cols)]\n", " show_images(Y, num_rows, num_cols, scale)" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### 9.1.1.1 翻转和裁剪" ] }, { "cell_type": "code", "execution_count": 5, "metadata": { "ExecuteTime": { "end_time": "2019-05-29T14:27:44.909361Z", "start_time": "2019-05-29T14:27:44.635038Z" } }, "outputs": [ { "data": { "image/svg+xml": [ "\n", "\n", "\n", "\n" ], "text/plain": [ "

" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" } ], "source": [ "apply(img, torchvision.transforms.RandomHorizontalFlip())" ] }, { "cell_type": "code", "execution_count": 6, "metadata": { "ExecuteTime": { "end_time": "2019-05-29T14:27:45.172398Z", "start_time": "2019-05-29T14:27:44.911747Z" } }, "outputs": [ { "data": { "image/svg+xml": [ "\n", "\n", "\n", "\n" ], "text/plain": [ "

" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" } ], "source": [ "apply(img, torchvision.transforms.RandomVerticalFlip())" ] }, { "cell_type": "code", "execution_count": 7, "metadata": { "ExecuteTime": { "end_time": "2019-05-29T14:27:45.452910Z", "start_time": "2019-05-29T14:27:45.174722Z" } }, "outputs": [ { "data": { "image/svg+xml": [ "\n", "\n", "\n", "\n" ], "text/plain": [ "

" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" } ], "source": [ "shape_aug = torchvision.transforms.RandomResizedCrop(200, scale=(0.1, 1), ratio=(0.5, 2))\n", "apply(img, shape_aug)" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### 9.1.1.2 变化颜色" ] }, { "cell_type": "code", "execution_count": 8, "metadata": { "ExecuteTime": { "end_time": "2019-05-29T14:27:45.711617Z", "start_time": "2019-05-29T14:27:45.454846Z" } }, "outputs": [ { "data": { "image/svg+xml": [ "\n", "\n", "\n", "\n" ], "text/plain": [ "

" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" } ], "source": [ "apply(img, torchvision.transforms.ColorJitter(brightness=0.5, contrast=0, saturation=0, hue=0))" ] }, { "cell_type": "code", "execution_count": 9, "metadata": { "ExecuteTime": { "end_time": "2019-05-29T14:27:46.056185Z", "start_time": "2019-05-29T14:27:45.713461Z" } }, "outputs": [ { "data": { "image/svg+xml": [ "\n", "\n", "\n", "\n" ], "text/plain": [ "

" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" } ], "source": [ "apply(img, torchvision.transforms.ColorJitter(brightness=0, contrast=0, saturation=0, hue=0.5))" ] }, { "cell_type": "code", "execution_count": 10, "metadata": { "ExecuteTime": { "end_time": "2019-05-29T14:27:46.349046Z", "start_time": "2019-05-29T14:27:46.057792Z" } }, "outputs": [ { "data": { "image/svg+xml": [ "\n", "\n", "\n", "\n" ], "text/plain": [ "

" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" } ], "source": [ "apply(img, torchvision.transforms.ColorJitter(brightness=0, contrast=0.5, saturation=0, hue=0))" ] }, { "cell_type": "code", "execution_count": 11, "metadata": { "ExecuteTime": { "end_time": "2019-05-29T14:27:46.794113Z", "start_time": "2019-05-29T14:27:46.351551Z" } }, "outputs": [ { "data": { "image/svg+xml": [ "\n", "\n", "\n", "\n" ], "text/plain": [ "

" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" } ], "source": [ "color_aug = torchvision.transforms.ColorJitter(brightness=0.5, contrast=0.5, saturation=0.5, hue=0.5)\n", "apply(img, color_aug)" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### 9.1.1.3 叠加多个图像增广方法" ] }, { "cell_type": "code", "execution_count": 12, "metadata": { "ExecuteTime": { "end_time": "2019-05-29T14:27:47.176935Z", "start_time": "2019-05-29T14:27:46.796137Z" } }, "outputs": [ { "data": { "image/svg+xml": [ "\n", "\n", "\n", "\n" ], "text/plain": [ "

" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" } ], "source": [ "augs = torchvision.transforms.Compose([\n", " torchvision.transforms.RandomHorizontalFlip(), color_aug, shape_aug])\n", "apply(img, augs)" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## 9.1.2 使用图像增广训练模型" ] }, { "cell_type": "code", "execution_count": 13, "metadata": { "ExecuteTime": { "end_time": "2019-05-29T14:27:50.991426Z", "start_time": "2019-05-29T14:27:47.179107Z" } }, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "Files already downloaded and verified\n" ] }, { "data": { "image/svg+xml": [ "\n", "\n", "\n", "\n" ], "text/plain": [ "

" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" } ], "source": [ "all_imges = torchvision.datasets.CIFAR10(train=True, root=\"~/Datasets/CIFAR\", download=True)\n", "# all_imges的每一个元素都是(image, label)\n", "show_images([all_imges[i][0] for i in range(32)], 4, 8, scale=0.8);" ] }, { "cell_type": "code", "execution_count": 14, "metadata": { "ExecuteTime": { "end_time": "2019-05-29T14:27:50.996901Z", "start_time": "2019-05-29T14:27:50.993693Z" } }, "outputs": [], "source": [ "flip_aug = torchvision.transforms.Compose([\n", " torchvision.transforms.RandomHorizontalFlip(),\n", " torchvision.transforms.ToTensor()])\n", "\n", "no_aug = torchvision.transforms.Compose([\n", " torchvision.transforms.ToTensor()])" ] }, { "cell_type": "code", "execution_count": 15, "metadata": { "ExecuteTime": { "end_time": "2019-05-29T14:27:51.003553Z", "start_time": "2019-05-29T14:27:50.998826Z" } }, "outputs": [], "source": [ "num_workers = 0 if sys.platform.startswith('win32') else 4\n", "def load_cifar10(is_train, augs, batch_size, root=\"~/Datasets/CIFAR\"):\n", " dataset = torchvision.datasets.CIFAR10(root=root, train=is_train, transform=augs, download=True)\n", " return DataLoader(dataset, batch_size=batch_size, shuffle=is_train, num_workers=num_workers)" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### 9.1.2.1 使用图像增广训练模型" ] }, { "cell_type": "code", "execution_count": 16, "metadata": { "ExecuteTime": { "end_time": "2019-05-29T14:27:51.010527Z", "start_time": "2019-05-29T14:27:51.005395Z" } }, "outputs": [], "source": [ "# 本函数已保存在d2lzh_pytorch包中方便以后使用\n", "def train(train_iter, test_iter, net, loss, optimizer, device, num_epochs):\n", " net = net.to(device)\n", " print(\"training on \", device)\n", " batch_count = 0\n", " for epoch in range(num_epochs):\n", " train_l_sum, train_acc_sum, n, start = 0.0, 0.0, 0, time.time()\n", " for X, y in train_iter:\n", " X = X.to(device)\n", " y = y.to(device)\n", " y_hat = net(X)\n", " l = loss(y_hat, y)\n", " optimizer.zero_grad()\n", " l.backward()\n", " optimizer.step()\n", " train_l_sum += l.cpu().item()\n", " train_acc_sum += (y_hat.argmax(dim=1) == y).sum().cpu().item()\n", " n += y.shape[0]\n", " batch_count += 1\n", " test_acc = d2l.evaluate_accuracy(test_iter, net)\n", " print('epoch %d, loss %.4f, train acc %.3f, test acc %.3f, time %.1f sec'\n", " % (epoch + 1, train_l_sum / batch_count, train_acc_sum / n, test_acc, time.time() - start))" ] }, { "cell_type": "code", "execution_count": 17, "metadata": { "ExecuteTime": { "end_time": "2019-05-29T14:27:51.016923Z", "start_time": "2019-05-29T14:27:51.012423Z" } }, "outputs": [], "source": [ "def train_with_data_aug(train_augs, test_augs, lr=0.001):\n", " batch_size, net = 256, d2l.resnet18(10)\n", " optimizer = torch.optim.Adam(net.parameters(), lr=lr)\n", " loss = torch.nn.CrossEntropyLoss()\n", " train_iter = load_cifar10(True, train_augs, batch_size)\n", " test_iter = load_cifar10(False, test_augs, batch_size)\n", " train(train_iter, test_iter, net, loss, optimizer, device, num_epochs=10)" ] }, { "cell_type": "code", "execution_count": 18, "metadata": { "ExecuteTime": { "end_time": "2019-05-29T14:48:32.502485Z", "start_time": "2019-05-29T14:27:51.018678Z" } }, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "Files already downloaded and verified\n", "Files already downloaded and verified\n", "training on cuda\n", "epoch 1, loss 1.3615, train acc 0.505, test acc 0.493, time 123.2 sec\n", "epoch 2, loss 0.5003, train acc 0.645, test acc 0.620, time 123.0 sec\n", "epoch 3, loss 0.2811, train acc 0.703, test acc 0.616, time 123.1 sec\n", "epoch 4, loss 0.1890, train acc 0.735, test acc 0.686, time 123.0 sec\n", "epoch 5, loss 0.1346, train acc 0.765, test acc 0.671, time 123.1 sec\n", "epoch 6, loss 0.1029, train acc 0.787, test acc 0.674, time 123.1 sec\n", "epoch 7, loss 0.0803, train acc 0.804, test acc 0.749, time 123.1 sec\n", "epoch 8, loss 0.0644, train acc 0.822, test acc 0.717, time 123.1 sec\n", "epoch 9, loss 0.0526, train acc 0.836, test acc 0.750, time 123.0 sec\n", "epoch 10, loss 0.0433, train acc 0.851, test acc 0.754, time 123.1 sec\n" ] } ], "source": [ "train_with_data_aug(flip_aug, no_aug)" ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [] } ], "metadata": { "kernelspec": { "display_name": "Python [conda env:py36]", "language": "python", "name": "conda-env-py36-py" }, "language_info": { "codemirror_mode": { "name": "ipython", "version": 3 }, "file_extension": ".py", "mimetype": "text/x-python", "name": "python", "nbconvert_exporter": "python", "pygments_lexer": "ipython3", "version": "3.6.8" }, "varInspector": { "cols": { "lenName": 16, "lenType": 16, "lenVar": 40 }, "kernels_config": { "python": { "delete_cmd_postfix": "", "delete_cmd_prefix": "del ", "library": "var_list.py", "varRefreshCmd": "print(var_dic_list())" }, "r": { "delete_cmd_postfix": ") ", "delete_cmd_prefix": "rm(", "library": "var_list.r", "varRefreshCmd": "cat(var_dic_list()) " } }, "types_to_exclude": [ "module", "function", "builtin_function_or_method", "instance", "_Feature" ], "window_display": false } }, "nbformat": 4, "nbformat_minor": 2 }