使用Albumentations进行数据增强

*Major*

已于 2024-01-25 15:02:46 修改

阅读量405

点赞数 7

文章标签：人工智能

于 2024-01-25 14:05:59 首次发布

本文链接：https://blog.csdn.net/qq_41375318/article/details/135843124

版权

使用Albumentations进行数据增强

https://juejin.cn/post/7080060840858091557

from major_utils import *
from PIL import Image
import albumentations as A


import cv2
import numpy as np


def scale_image_max(image):
    # 将图像转换为浮点数格式
    image_float = image.astype(np.float32)

    # 找到像素值范围
    min_val, max_val = np.min(image_float), np.max(image_float)

    # 将像素值缩放到0-255之间
    image_normalized = cv2.normalize(image_float, None, alpha=0, beta=255, norm_type=cv2.NORM_MINMAX, dtype=cv2.CV_8U)

    # 返回归一化后的图像
    return image_normalized

img_size = [1024, 1024]

data_transforms =  {
    "train": A.Compose([
        # 改变图像大小
        A.Resize(*img_size, interpolation=cv2.INTER_NEAREST),
        # 水平翻转
        A.HorizontalFlip(p=0.5),
        # 垂直反转
        A.VerticalFlip(p=0.5),
        # 转置
        A.Transpose(always_apply=False, p=0.3),
        # 平移、缩放、旋转
        A.ShiftScaleRotate(shift_limit=0.05, scale_limit=0.1, rotate_limit=45, interpolation=1, border_mode=4, value=None, mask_value=None, always_apply=False, p=1),
        # 随机对比度增强
        A.RandomBrightnessContrast(brightness_limit=(-0.1,0.1), contrast_limit=(0.0,0.2),p=0),
        # 弹性变换
        A.ElasticTransform(alpha=5, sigma=50, alpha_affine=50, interpolation=1, border_mode=4, always_apply=False, p=0.3)
        # 标准化图像
        # A.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]),
    ], p=1.0),

    "valid": A.Compose([
        A.Resize(*img_size, interpolation=cv2.INTER_NEAREST),
    ], p=1.0)
}

# 定义数据增强流程
# transform = Compose([
#     RandomCrop(height=224, width=224),  # 随机裁剪到224x224大小
#     RandomRotation(degrees=10),  # 随机旋转10度
#     Resize(height=224, width=224),  # 重新调整图像大小（可选，如果需要的话）
#     Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]),  # 标准化图像（可选，如果需要的话）
# ], p=1)  # p=1 表示这些操作都会被应用到一个图像上

for i in range(100):
    data_transform = data_transforms['train']

    img = cv2.imread(r"images\0200.tif", cv2.IMREAD_UNCHANGED)
    scaled_image = scale_image_max(img)
    scaled_image2 = scaled_image * 2.1 - 90
    # scaled_image2 = scaled_image2.astype(np.uint8)
    img = np.clip(scaled_image2, 0, 255).astype(np.uint8)



    msk = cv2.imread(r"labels\0200.tif", cv2.IMREAD_UNCHANGED)
    data = data_transform(image=img, mask=msk)
    img = data['image']
    msk = data['mask']

    cv2.imshow('img', img)
    cv2.imshow('msk', msk)
    cv2.waitKey(0)
    cv2.destroyAllWindows()