李沐d2l(十一)--目标检测

一、概念

图片分类是给出指定图片，判断里面的主体类别，往往一张图片只有一个主体。而目标检测是要把含有多个主体的图片，识别出用户感兴趣的主体，一般通过方框(边缘)标注。

边缘框

一个边缘框可以通过4个数字定义

1. 左上x，左上y，右下x，右下y
2. 左上x，左上y，宽，高

二、代码

1 获取待检测图片

d2l.set_figsize()
img = d2l.plt.imread("catdog.jpg")
d2l.plt.imshow(img)
d2l.plt.show()

2 定义两种边缘框表示之间的转换函数

def box_corner_to_center(boxes):
    """从（左上，右下）转换到（中间，宽度，高度）"""
    x1, y1, x2, y2 = boxes[:, 0], boxes[:, 1], boxes[:, 2], boxes[:, 3]
    cx = (x1 + x2) / 2
    cy = (y1 + y2) / 2
    w = x2 - x1
    h = y2 - y1
    boxes = torch.stack((cx, cy, w, h), axis=-1)
    return boxes

def box_center_to_corner(boxes):
    """从（中间，宽度，高度）转换到（左上，右下）"""
    cx, cy, w, h = boxes[:, 0], boxes[:, 1], boxes[:, 2], boxes[:, 3]
    x1 = cx - 0.5 * w
    y1 = cy - 0.5 * h
    x2 = cx + 0.5 * w
    y2 = cy + 0.5 * h
    boxes = torch.stack((x1, y1, x2, y2), axis=-1)
    return boxes

3 定义图像中狗和猫的边界框

dog_bbox, cat_bbox = [60.0, 45.0, 378.0, 516.0], [400.0, 112.0, 655.0, 493.0]

boxes = torch.tensor((dog_bbox, cat_bbox))

4 画出边界框

def bbox_to_rect(bbox, color):
    return d2l.plt.Rectangle(xy=(bbox[0], bbox[1]), width=bbox[2] - bbox[0],
                             height=bbox[3] - bbox[1], fill=False,
                             edgecolor=color, linewidth=2)

fig = d2l.plt.imshow(img)
fig.axes.add_patch(bbox_to_rect(dog_bbox, 'blue'))
fig.axes.add_patch(bbox_to_rect(cat_bbox, 'red'))
d2l.plt.show()

三、目标检测数据集

1 下载数据集

import os
import pandas as pd
import torch
import torchvision
from d2l import torch as d2l
os.environ["KMP_DUPLICATE_LIB_OK"]="TRUE"

d2l.DATA_HUB['banana-detection'] = (
    d2l.DATA_URL + 'banana-detection.zip',
    '5de26c8fce5ccdea9f91267273464dc968d20d72')

2 读取数据集

def read_data_bananas(is_train=True):
    """读取香蕉检测数据集中的图像和标签。"""
    data_dir = d2l.download_extract('banana-detection')
    csv_fname = os.path.join(data_dir,
                             'bananas_train' if is_train else 'bananas_val',
                             'label.csv')
    csv_data = pd.read_csv(csv_fname)
    csv_data = csv_data.set_index('img_name')
    images, targets = [], []
    for img_name, target in csv_data.iterrows():
        images.append(
            torchvision.io.read_image(
                os.path.join(data_dir,
                             'bananas_train' if is_train else 'bananas_val',
                             'images', f'{img_name}')))
        targets.append(list(target))
    return images, torch.tensor(targets).unsqueeze(1) / 256

3 自定义Dataset实例

class BananasDataset(torch.utils.data.Dataset):
    """一个用于加载香蕉检测数据集的自定义数据集。"""
    def __init__(self, is_train):
        self.features, self.labels = read_data_bananas(is_train)
        print('read ' + str(len(self.features)) + (
            f' training examples' if is_train else f' validation examples'))

    def __getitem__(self, idx):
        return (self.features[idx].float(), self.labels[idx])

    def __len__(self):
        return len(self.features)

4 返回训练集合测试集加载器实例

def load_data_bananas(batch_size):
    """加载香蕉检测数据集。"""
    train_iter = torch.utils.data.DataLoader(BananasDataset(is_train=True),
                                             batch_size, shuffle=True)
    val_iter = torch.utils.data.DataLoader(BananasDataset(is_train=False),
                                           batch_size)
    return train_iter, val_iter

5 读取一个小批量

batch_size, edge_size = 32, 256
train_iter, _ = load_data_bananas(batch_size)
batch = next(iter(train_iter))

6 展示检测结果

imgs = (batch[0][0:10].permute(0, 2, 3, 1)) / 255
axes = d2l.show_images(imgs, 2, 5, scale=2)
for ax, label in zip(axes, batch[1][0:10]):
    d2l.show_bboxes(ax, [label[0][1:5] * edge_size], colors=['w'])
d2l.plt.show()