企业官网建设流程全解析

织梦网站首页,网站开发什么课程,专项培训网站建设方案,网站模板 哪家好前言#xff1a; 书接上回#xff0c;我们在上一章详细讲解了视觉学习中的数据预处理以及其相应的label的格式类型互转的特性#xff0c;之后又进一步对数据增强的方法进行进一步的总结#xff0c;接下来这一章是延续上一章节的数据增强方法进行拓展。 视觉opencv学习笔记Ⅴ…前言书接上回我们在上一章详细讲解了视觉学习中的数据预处理以及其相应的label的格式类型互转的特性之后又进一步对数据增强的方法进行进一步的总结接下来这一章是延续上一章节的数据增强方法进行拓展。视觉opencv学习笔记Ⅴ-数据增强(1)-CSDN博客1.像素变换类数据增强锦集⭐1.BGR-RGB/BGR-灰度图核心逻辑BGR - RGB-灰度图import cv2 import os import numpy as np # 核心函数BGR转灰度图 def bgr2gray(img_bgr: np.ndarray, to_3ch: bool True) - np.ndarray: 极简核心BGR转灰度图可选转为3通道适配模型输入 :param img_bgr: OpenCV读取的BGR格式图像H,W,C :param to_3ch: 是否转为3通道True模型兼容False单通道 :return: 灰度图单/3通道 img_gray cv2.cvtColor(img_bgr, cv2.COLOR_BGR2GRAY) # 3通道转换重复单通道数据 if to_3ch: return cv2.cvtColor(img_gray, cv2.COLOR_GRAY2BGR) return img_gray # 核心调用示例 if __name__ __main__: # 1. 配置路径仅改这里 IMG_PATH rC:\Users\Excub\workspace\Pytorch\12.08_cv\perspective_data\ok.jpg SAVE_DIR rC:\Users\Excub\workspace\Pytorch\12.08_cv\perspective_data\pixel_aug_results # 2. 加载原图极简版 img_bgr cv2.imread(IMG_PATH) if img_bgr is None: raise FileNotFoundError(f无法读取图片{IMG_PATH}) # 3. 核心调用转灰度图3通道适配模型 img_gray_3ch bgr2gray(img_bgr, to_3chTrue) # 可选转单通道灰度图 # img_gray_1ch bgr2gray(img_bgr, to_3chFalse) # 4. 保存结果仅保留核心保存 os.makedirs(SAVE_DIR, exist_okTrue) img_basename os.path.splitext(os.path.basename(IMG_PATH))[0] cv2.imwrite(f{SAVE_DIR}/{img_basename}_gray_3ch.jpg, img_gray_3ch) # cv2.imwrite(f{SAVE_DIR}/{img_basename}_gray_1ch.jpg, img_gray_1ch) # 单通道保存 print(f转灰度图完成保存路径{SAVE_DIR}/{img_basename}_gray_3ch.jpg) print(核心要点) print(1. to_3chTrue灰度图转为3通道可直接输入要求3通道的模型) print(2. to_3chFalse输出单通道灰度图仅保留亮度信息) print(3. 无色彩失真灰度转换仅丢失色度信息亮度特征完整保留。)2.全局直方图均衡化cv2.equalizeHist(gray_img) # 仅支持单通道灰度/Y通道核心代码def global_histogram_equalization(img_bgr: np.ndarray) - tuple[np.ndarray, np.ndarray]: 全局直方图均衡化工业级实现区分灰度/彩色图 Args: img_bgr: BGR格式原图 Returns: img_gray_eq: 灰度图均衡化结果3通道 img_color_eq: 彩色图均衡化结果仅均衡化亮度通道保留色彩 # 1. 灰度图直方图均衡化 img_gray cv2.cvtColor(img_bgr, cv2.COLOR_BGR2GRAY) img_gray_eq cv2.equalizeHist(img_gray) # 核心API全局直方图均衡化 img_gray_eq_3ch cv2.cvtColor(img_gray_eq, cv2.COLOR_GRAY2BGR) # 转3通道 print(完成灰度图全局直方图均衡化提升整体对比度) # 2. 彩色图直方图均衡化关键仅处理亮度通道 # 彩色图不能直接均衡化BGR通道会色彩失真需转YCrCb空间Y亮度Cr/Cb色彩 img_ycrcb cv2.cvtColor(img_bgr, cv2.COLOR_BGR2YCrCb) # 仅对亮度通道Y做均衡化 img_ycrcb[:, :, 0] cv2.equalizeHist(img_ycrcb[:, :, 0]) # 转回BGR空间 img_color_eq cv2.cvtColor(img_ycrcb, cv2.COLOR_YCrCb2BGR) print(完成彩色图全局直方图均衡化仅均衡化亮度通道无色彩失真) return img_gray_eq_3ch, img_color_eq3.自适应均衡化cv2.createCLAHE()区别于全局均衡化的cv2.equalizeHist()import cv2 import numpy as np def adaptive_histogram_equalization(img_bgr: np.ndarray) - tuple[np.ndarray, np.ndarray]: 自适应区域直方图均衡化核心方法 关键参数 - clipLimit2.0限制对比度避免过度增强放大噪声 - tileGridSize(8,8)分块大小将图像分为8×8的小区域分别均衡化 # 1. 灰度图自适应均衡化 img_gray cv2.cvtColor(img_bgr, cv2.COLOR_BGR2GRAY) # 创建CLAHE对象核心API clahe cv2.createCLAHE(clipLimit2.0, tileGridSize(8, 8)) img_gray_clahe clahe.apply(img_gray) # 应用自适应均衡化 img_gray_clahe_3ch cv2.cvtColor(img_gray_clahe, cv2.COLOR_GRAY2BGR) # 2. 彩色图自适应均衡化仅亮度通道 img_ycrcb cv2.cvtColor(img_bgr, cv2.COLOR_BGR2YCrCb) img_ycrcb[:, :, 0] clahe.apply(img_ycrcb[:, :, 0]) # 仅处理亮度通道 img_color_clahe cv2.cvtColor(img_ycrcb, cv2.COLOR_YCrCb2BGR) return img_gray_clahe_3ch, img_color_clahe # ------------------- 核心调用示例 ------------------- if __name__ __main__: # 1. 加载原图 IMG_PATH rC:\Users\Excub\workspace\Pytorch\12.08_cv\perspective_data\ok.jpg img_origin cv2.imread(IMG_PATH, cv2.IMREAD_COLOR) # 2. 核心调用自适应直方图均衡化 img_gray_clahe, img_color_clahe adaptive_histogram_equalization(img_origin) # 3. 保存结果可选用于对比 SAVE_DIR rC:\Users\Excub\workspace\Pytorch\12.08_cv\perspective_data\pixel_aug_results cv2.imwrite(f{SAVE_DIR}/ok_gray_adaptive_hist_eq.jpg, img_gray_clahe) cv2.imwrite(f{SAVE_DIR}/ok_color_adaptive_hist_eq.jpg, img_color_clahe) print(自适应直方图均衡化完成) print(关键差异) print(1. 对比全局均衡化局部明暗不均区域如部分逆光会更自然无过曝/噪声放大) print(2. clipLimit越大对比度增强越明显但易放大噪声建议2.0工业常用) print(3. tileGridSize越小分块越细局部适配越好但计算量略增8×8为最优平衡。)4.随机调节亮度random_brightness随机调整亮度import cv2 import numpy as np import random def random_brightness(img_bgr: np.ndarray, brightness_range(0.5, 1.5)) - np.ndarray: 随机调整亮度核心方法 关键约束 - 亮度因子限制在 0.5~1.5避免过暗/过曝 - 像素值强制 clip 到 0~255防止异常值 # 随机生成亮度因子0.5~1.5 倍 brightness random.uniform(*brightness_range) # 转为 float 计算避免 uint8 溢出 img_float img_bgr.astype(np.float32) # 调整亮度所有通道同比例调整保留色彩 img_bright img_float * brightness # 限制像素值在 0~255 之间核心约束 img_bright np.clip(img_bright, 0, 255).astype(np.uint8) return img_bright # ------------------- 核心调用示例 ------------------- if __name__ __main__: # 1. 加载原图 IMG_PATH rC:\Users\Excub\workspace\Pytorch\12.08_cv\perspective_data\ok.jpg img_origin cv2.imread(IMG_PATH, cv2.IMREAD_COLOR) # 2. 核心调用随机调整亮度0.5~1.5 倍 img_bright random_brightness(img_origin, brightness_range(0.5, 1.5)) # 3. 保存结果用于对比 SAVE_DIR rC:\Users\Excub\workspace\Pytorch\12.08_cv\perspective_data\pixel_aug_results cv2.imwrite(f{SAVE_DIR}/ok_random_brightness.jpg, img_bright) print(随机亮度调整完成) print(关键要点) print(1. 亮度因子建议 0.5~1.50.3 过暗丢失细节1.8 过曝) print(2. 所有通道同比例调整避免色彩偏移如仅调R通道导致偏色) print(3. 必须 clip 像素值uint8 范围 0~255溢出会出现伪色/黑块) print(4. 适用场景模拟强光/弱光下的工业目标如车间灯光明暗变化。)5.图像像素取反核心调用方式# 仅执行像素取反直接使用 img_bgr cv2.imread(IMG_PATH) img_inverted img_bitwise_not(img_bgr) cv2.imwrite(ok_bitwise_not.jpg, img_inverted)import cv2 import os import numpy as np # 核心函数图像像素取反反色 def img_bitwise_not(img_bgr: np.ndarray) - np.ndarray: 极简核心OpenCV BGR图像像素取反反色 :param img_bgr: OpenCV读取的BGR格式图像H,W,C :return: 反色后的BGR图像 # 核心APIcv2.bitwise_not 直接对像素取反uint8范围0~255等价于 255 - 像素值 img_inverted cv2.bitwise_not(img_bgr) return img_inverted # 极简加载/保存标签取反无需修改标签仅拷贝 def copy_yolo_label(src_label_path: str, dst_label_path: str): 复制原标签取反不改变目标框无需修正 if os.path.exists(src_label_path): import shutil shutil.copy(src_label_path, dst_label_path) print(f标签已拷贝{dst_label_path}) # 核心调用示例 if __name__ __main__: # 1. 配置路径仅改这里 IMG_PATH rC:\Users\Excub\workspace\Pytorch\12.08_cv\perspective_data\ok.jpg LABEL_PATH rC:\Users\Excub\workspace\Pytorch\12.08_cv\perspective_data\ok.txt # 可选原YOLO标签 SAVE_DIR rC:\Users\Excub\workspace\Pytorch\12.08_cv\perspective_data\pixel_aug_results # 2. 加载原图极简版 img_bgr cv2.imread(IMG_PATH) if img_bgr is None: raise FileNotFoundError(f无法读取图片{IMG_PATH}) # 3. 核心调用像素取反 img_inverted img_bitwise_not(img_bgr) # 4. 保存结果 os.makedirs(SAVE_DIR, exist_okTrue) img_basename os.path.splitext(os.path.basename(IMG_PATH))[0] cv2.imwrite(f{SAVE_DIR}/{img_basename}_bitwise_not.jpg, img_inverted) # 5. 拷贝标签取反不改变目标框直接复用原标签 if os.path.exists(LABEL_PATH): copy_yolo_label(LABEL_PATH, f{SAVE_DIR}/{img_basename}_bitwise_not.txt) print(f像素取反完成保存路径{SAVE_DIR}/{img_basename}_bitwise_not.jpg) print(核心要点) print(1. 核心效果像素值取反如黑色→白色、红色→青色增强目标与背景对比度) print(2. 标签处理无需修正YOLO框仅像素值变化目标位置/大小不变) print(3. 适用场景深色背景下的浅色目标、黑白对比强烈的工业场景如印刷字符) print(4. 无参数调整纯像素值反转无超参需调优。)6.减去像素均值subtract_mean核心调用# 自动计算全局均值并减去 img_bgr cv2.imread(IMG_PATH) img_sub_mean subtract_mean(img_bgr, meanNone) cv2.imwrite(ok_sub_mean.jpg, img_sub_mean)作用像素级归一化操作1. 消除全局光照偏移的影响2. 降低数据分布差异提升模型泛化性场景1. 光照不稳定的工业 / 户外视觉场景核心场景工业检测车间零件 / 包装盒拍摄灯光明暗、角度导致光照不均户外视觉交通摄像头、无人机图像早晚 / 阴天 / 晴天光照差异大安防监控夜间 / 白天的行人 / 车辆检测全局亮度波动大。2. 基于预训练模型的迁移学习必用场景只要你在分类、检测、分割任务中复用「ImageNet 预训练权重」必须将输入图像减去预训练时的固定均值如 ImageNet 均值否则模型的底层特征提取如边缘、纹理会完全偏离预期。例用 ResNet50 做工业零件缺陷分类第一步预处理就是subtract_mean(img, mean[103.939, 116.779, 123.68])使用必须要做的事情必须做clip约束减去均值后可能出现负数像素如暗区像素值 5均值 10结果 - 5需用np.clip(img_sub_mean, 0, 255)限制到 0~255否则图像会出现伪色、黑块等异常均值选择要匹配场景迁移学习用预训练模型的固定均值如 ImageNet自有数据集训练计算数据集的全局均值所有图片的通道均值而非单张图片均值代码中np.mean(img_float, axis(0,1))是单张均值批量训练需先算全量均值仅像素值变化标签无需修改和像素取反、亮度调整一样减去均值不改变目标的位置 / 大小YOLO 标签直接拷贝即可。#!/usr/bin/env python # -*- coding: utf-8 -*- Project Pytorch File subtract_mean.py IDE PyCharm Author wjj Date 2025/12/14 18:56 Description: import cv2 import os import numpy as np # 核心函数减去像素均值 def subtract_mean(img_bgr: np.ndarray, mean: list None) - np.ndarray: 极简核心图像减去像素均值全局/自定义clip避免像素值异常 :param img_bgr: OpenCV读取的BGR图像H,W,C :param mean: 自定义均值 [B, G, R]None则计算图像全局均值 :return: 减去均值后的BGR图像uint8 # 转为float计算避免uint8溢出负数/超界 img_float img_bgr.astype(np.float32) # 计算/使用均值 if mean is None: mean np.mean(img_float, axis(0, 1)) # 按通道计算全局均值 print(f自动计算图像全局均值B/G/R{mean.round(2)}) else: mean np.array(mean, dtypenp.float32) # 核心操作减去均值 约束像素值到0~255 img_sub_mean img_float - mean img_sub_mean np.clip(img_sub_mean, 0, 255) # 避免负数/超255导致图像异常 # 转回uint8符合OpenCV保存/模型输入要求 return img_sub_mean.astype(np.uint8) # 极简标签处理直接拷贝 def copy_yolo_label(src_label_path: str, dst_label_path: str): 拷贝原标签均值减法不改变目标框 if os.path.exists(src_label_path): import shutil shutil.copy(src_label_path, dst_label_path) # 核心调用示例 if __name__ __main__: # 1. 配置路径 IMG_PATH rC:\Users\Excub\workspace\Pytorch\12.08_cv\perspective_data\ok.jpg LABEL_PATH rC:\Users\Excub\workspace\Pytorch\12.08_cv\perspective_data\ok.txt # 可选 SAVE_DIR rC:\Users\Excub\workspace\Pytorch\12.08_cv\perspective_data\pixel_aug_results # 2. 加载原图 img_bgr cv2.imread(IMG_PATH) if img_bgr is None: raise FileNotFoundError(f无法读取图片{IMG_PATH}) # 3. 核心调用两种方式可选 # 方式1自动计算全局均值 img_sub_global_mean subtract_mean(img_bgr, meanNone) # 方式2使用自定义均值如ImageNet均值 [103.939, 116.779, 123.68]适配预训练模型 # img_sub_custom_mean subtract_mean(img_bgr, mean[103.939, 116.779, 123.68]) # 4. 保存结果 os.makedirs(SAVE_DIR, exist_okTrue) img_basename os.path.splitext(os.path.basename(IMG_PATH))[0] cv2.imwrite(f{SAVE_DIR}/{img_basename}_sub_global_mean.jpg, img_sub_global_mean) # cv2.imwrite(f{SAVE_DIR}/{img_basename}_sub_custom_mean.jpg, img_sub_custom_mean) # 5. 拷贝标签无需修正 if os.path.exists(LABEL_PATH): copy_yolo_label(LABEL_PATH, f{SAVE_DIR}/{img_basename}_sub_global_mean.txt) print(f减去均值完成保存路径{SAVE_DIR}/{img_basename}_sub_global_mean.jpg) print(核心要点) print(1. 核心作用降低光照全局偏移影响提升模型泛化性) print(2. 关键约束必须clip到0~255否则负数像素会导致图像偏色/异常) print(3. 均值选择自定义均值如ImageNet适配预训练模型全局均值适配自有数据) print(4. 标签处理仅像素值偏移目标框不变直接拷贝原标签。)7.图像锐化增强拉普拉斯 / 高斯 / 均值核心调用# 仅执行高斯锐化最常用效果柔和 img_bgr cv2.imread(IMG_PATH) img_sharpen image_sharpen(img_bgr, sharpen_typegaussian, alpha0.5) cv2.imwrite(ok_gaussian_sharpen.jpg, img_sharpen)其本质作用为利用 “模糊 / 边缘提取” 的结果反向增强细节。其与去噪的三种模糊效果完全相反。可以把图像想象成 “一张有纹理的纸”模糊均值 / 高斯用手把纸揉平纹理细节变浅弱化锐化均值 / 高斯把揉平的纸和原纸对比找出 “变浅的纹理差”再用笔把这些纹理描深强化拉普拉斯锐化直接找到纸的 “轮廓线”边缘用笔把轮廓线描粗强强化。场景需求选模糊均值 / 高斯选锐化均值 / 高斯 / 拉普拉斯图像有颗粒噪声需降噪✅ 均值模糊❌ 锐化会放大噪声低分辨率图需平滑轮廓✅ 高斯模糊❌ 锐化会放大像素块模糊的零件划痕需清晰❌ 模糊会更糊✅ 拉普拉斯锐化强边缘印刷字边缘模糊需柔和增强❌ 模糊会更糊✅ 高斯锐化自然无伪影纹理对比度低需强化❌ 模糊会降低对比度✅ 均值锐化提升纹理对比import cv2 import os import numpy as np # 核心函数多类型图像锐化 def image_sharpen(img_bgr: np.ndarray, sharpen_type: str laplacian, alpha: float 0.5) - np.ndarray: 极简核心拉普拉斯/高斯/均值锐化增强边缘/纹理 :param img_bgr: OpenCV读取的BGR图像H,W,C :param sharpen_type: 锐化类型 - laplacian/gaussian/mean :param alpha: 锐化强度0.3~0.7为宜过强产生伪影 :return: 锐化后的BGR图像uint8 # 转为float计算避免uint8溢出 img_float img_bgr.astype(np.float32) # 1. 生成模糊/边缘图不同锐化类型的核心差异 if sharpen_type laplacian: # 拉普拉斯锐化提取边缘后叠加到原图 laplacian cv2.Laplacian(img_float, cv2.CV_32F, ksize3) # 3×3核提取边缘 img_sharpen img_float alpha * laplacian elif sharpen_type gaussian: # 高斯锐化原图 - 高斯模糊图 → 增强细节 img_blur cv2.GaussianBlur(img_float, (3, 3), 1.0) # 3×3高斯核sigma1 img_sharpen img_float alpha * (img_float - img_blur) elif sharpen_type mean: # 均值锐化原图 - 均值模糊图 → 增强对比度 img_blur cv2.blur(img_float, (3, 3)) # 3×3均值核 img_sharpen img_float alpha * (img_float - img_blur) else: raise ValueError(f无效锐化类型{sharpen_type}可选laplacian/gaussian/mean) # 2. 约束像素值到0~255避免伪影/溢出 img_sharpen np.clip(img_sharpen, 0, 255).astype(np.uint8) return img_sharpen # 极简标签处理直接拷贝 def copy_yolo_label(src_label_path: str, dst_label_path: str): 拷贝原标签锐化不改变目标框 if os.path.exists(src_label_path): import shutil shutil.copy(src_label_path, dst_label_path) # 核心调用示例 if __name__ __main__: # 1. 配置路径 IMG_PATH rC:\Users\Excub\workspace\Pytorch\12.08_cv\jjy.jpg LABEL_PATH rC:\Users\Excub\workspace\Pytorch\12.08_cv\perspective_data\ok.txt # 可选 SAVE_DIR rC:\Users\Excub\workspace\Pytorch\12.08_cv\out # 2. 加载原图 img_bgr cv2.imread(IMG_PATH) if img_bgr is None: raise FileNotFoundError(f无法读取图片{IMG_PATH}) # 3. 核心调用三种锐化按需选一种 img_laplacian image_sharpen(img_bgr, sharpen_typelaplacian, alpha0.5) # 拉普拉斯边缘最强 img_gaussian image_sharpen(img_bgr, sharpen_typegaussian, alpha0.5) # 高斯最柔和 img_mean image_sharpen(img_bgr, sharpen_typemean, alpha0.5) # 均值对比度增强 # 4. 保存结果 os.makedirs(SAVE_DIR, exist_okTrue) img_basename os.path.splitext(os.path.basename(IMG_PATH))[0] cv2.imwrite(f{SAVE_DIR}/{img_basename}_laplacian_sharpen.jpg, img_laplacian) cv2.imwrite(f{SAVE_DIR}/{img_basename}_gaussian_sharpen.jpg, img_gaussian) cv2.imwrite(f{SAVE_DIR}/{img_basename}_mean_sharpen.jpg, img_mean) # 5. 拷贝标签无需修正 if os.path.exists(LABEL_PATH): copy_yolo_label(LABEL_PATH, f{SAVE_DIR}/{img_basename}_laplacian_sharpen.txt) print(三种锐化完成保存路径, SAVE_DIR) print(核心要点) print(1. 锐化强度alpha0.3~0.7为宜0.8易产生锯齿/伪影0.2效果不明显) print(2. 类型选择) print( - laplacian边缘增强最强适合模糊的目标边缘如远距离小目标) print( - gaussian锐化最柔和适合低分辨率/运动模糊图像无明显伪影) print( - mean对比度增强适合纹理模糊的目标如印刷字符/零件纹理) print(3. 适用场景低分辨率、运动模糊、边缘模糊的工业目标如包装盒文字、零件划痕) print(4. 标签处理仅增强纹理目标框不变直接拷贝原标签。)8.Sobel 边缘融合增强本质Sobel 边缘提取 边缘融合到原图核心逻辑强化目标边缘特征# 仅执行双方向Sobel边缘融合最常用 img_bgr cv2.imread(IMG_PATH) img_sobel sobel_edge_to_image(img_bgr, edge_dirboth, alpha0.5) cv2.imwrite(ok_sobel_both.jpg, img_sobel)Sobel 算子是 “边缘提取刀”高斯算子是 “降噪洗菜盆”二者都是底层工具Sobel 边缘融合增强是 “用刀 盆可选做一道边缘更清晰的菜”是上层应用流程#!/usr/bin/env python # -*- coding: utf-8 -*- Project Pytorch File sobel_.py IDE PyCharm Author wjj Date 2025/12/14 19:23 Description: import cv2 import os import numpy as np # 核心函数Sobel边缘提取融合到原图 def sobel_edge_to_image(img_bgr: np.ndarray, edge_dir: str both, alpha: float 0.5) - np.ndarray: 极简核心Sobel边缘提取 边缘图融合到原图强化边缘特征 :param img_bgr: OpenCV读取的BGR图像H,W,C :param edge_dir: 边缘检测方向 - x垂直/y水平/both双方向 :param alpha: 边缘融合强度0.3~0.8为宜过强掩盖原图细节 :return: 融合边缘后的BGR图像uint8 # 1. 转为灰度图Sobel仅支持单通道 img_gray cv2.cvtColor(img_bgr, cv2.COLOR_BGR2GRAY).astype(np.float32) # 2. Sobel边缘提取核心APIcv2.Sobel一阶导数边缘检测 if edge_dir x: # x方向检测垂直边缘如文字竖线、零件竖边 sobel_edge cv2.Sobel(img_gray, cv2.CV_32F, dx1, dy0, ksize3) elif edge_dir y: # y方向检测水平边缘如文字横线、零件横边 sobel_edge cv2.Sobel(img_gray, cv2.CV_32F, dx0, dy1, ksize3) elif edge_dir both: # 双方向合并x/y边缘最常用完整边缘 sobel_x cv2.Sobel(img_gray, cv2.CV_32F, dx1, dy0, ksize3) sobel_y cv2.Sobel(img_gray, cv2.CV_32F, dx0, dy1, ksize3) sobel_edge cv2.addWeighted(sobel_x, 0.5, sobel_y, 0.5, 0) else: raise ValueError(f无效边缘方向{edge_dir}可选x/y/both) # 3. 边缘图归一化0~255 转为3通道匹配原图 sobel_edge np.clip(cv2.normalize(sobel_edge, None, 0, 255, cv2.NORM_MINMAX), 0, 255) sobel_edge_3ch cv2.cvtColor(sobel_edge.astype(np.uint8), cv2.COLOR_GRAY2BGR) # 4. 边缘图融合到原图核心原图为主边缘图为辅 img_float img_bgr.astype(np.float32) edge_float sobel_edge_3ch.astype(np.float32) img_fused (1 - alpha) * img_float alpha * edge_float # 加权融合 img_fused np.clip(img_fused, 0, 255).astype(np.uint8) return img_fused # 极简标签处理直接拷贝 def copy_yolo_label(src_label_path: str, dst_label_path: str): 拷贝原标签边缘融合不改变目标框 if os.path.exists(src_label_path): import shutil shutil.copy(src_label_path, dst_label_path) # 核心调用示例 if __name__ __main__: # 1. 配置路径 IMG_PATH rC:\Users\Excub\workspace\Pytorch\12.08_cv\jjy.jpg LABEL_PATH rC:\Users\Excub\workspace\Pytorch\12.08_cv\perspective_data\ok.txt # 可选 SAVE_DIR rC:\Users\Excub\workspace\Pytorch\12.08_cv\out\sobel # 2. 加载原图 img_bgr cv2.imread(IMG_PATH) if img_bgr is None: raise FileNotFoundError(f无法读取图片{IMG_PATH}) # 3. 核心调用按需选方向 img_sobel_x sobel_edge_to_image(img_bgr, edge_dirx, alpha0.5) # 垂直边缘 img_sobel_y sobel_edge_to_image(img_bgr, edge_diry, alpha0.5) # 水平边缘 img_sobel_both sobel_edge_to_image(img_bgr, edge_dirboth, alpha0.5) # 双方向最常用 # 4. 保存结果 os.makedirs(SAVE_DIR, exist_okTrue) img_basename os.path.splitext(os.path.basename(IMG_PATH))[0] cv2.imwrite(f{SAVE_DIR}/{img_basename}_sobel_x.jpg, img_sobel_x) cv2.imwrite(f{SAVE_DIR}/{img_basename}_sobel_y.jpg, img_sobel_y) cv2.imwrite(f{SAVE_DIR}/{img_basename}_sobel_both.jpg, img_sobel_both) # 5. 拷贝标签无需修正 if os.path.exists(LABEL_PATH): copy_yolo_label(LABEL_PATH, f{SAVE_DIR}/{img_basename}_sobel_both.txt) print(Sobel边缘融合完成保存路径, SAVE_DIR) print(核心要点) print(1. 边缘方向选择) print( - x方向强化垂直边缘如包装盒竖边、文字竖线) print( - y方向强化水平边缘如包装盒横边、文字横线) print( - both完整强化所有边缘工业场景首选) print(2. 强度alpha0.3~0.8为宜0.8会掩盖原图纹理0.3边缘增强不明显) print(3. 适用场景边缘模糊的工业目标如低分辨率零件边缘、印刷字符轮廓) print(4. 核心差异Sobel是一阶导数边缘柔和拉普拉斯是二阶锐利Sobel更适合工业模糊图。)2.图像融合 / 拼接类数据增强锦集1.MixUp 数据增强核心将两张图按照一定的权重进行融合为一张图核心代码# 仅执行MixUp融合λ0.5 img1 cv2.imread(IMG1_PATH) img2 cv2.imread(IMG2_PATH) bboxes1 load_yolo_bboxes(LABEL1_PATH) bboxes2 load_yolo_bboxes(LABEL2_PATH) img_mix, bboxes_mix mix_up_images(img1, img2, bboxes1, bboxes2, lambda_0.5)#!/usr/bin/env python # -*- coding: utf-8 -*- Project Pytorch File mixup.py IDE PyCharm Author wjj Date 2025/12/14 19:41 Description:mixup新增带Box可视化功能 import cv2 import os import numpy as np import random # 核心函数MixUp图像标签融合 def mix_up_images(img1: np.ndarray, img2: np.ndarray, bboxes1: np.ndarray, bboxes2: np.ndarray, lambda_: float None) - tuple[np.ndarray, np.ndarray]: 极简核心两张图像MixUp融合 YOLO标签合并检测场景 :param img1: 第一张BGR图像H,W,C :param img2: 第二张BGR图像需和img1同尺寸 :param bboxes1: 第一张图YOLO归一化框 [x,y,w,h]shape(n,4) :param bboxes2: 第二张图YOLO归一化框 [x,y,w,h]shape(m,4) :param lambda_: 融合系数0~1None则随机生成0.3~0.7 :return: mix_up后的图像、合并后的YOLO框nm,4 # 1. 确保两张图尺寸一致MixUp前提 if img1.shape ! img2.shape: img2 cv2.resize(img2, (img1.shape[1], img1.shape[0])) # 2. 随机生成融合系数λ0.3~0.7避免某张图占比过高 if lambda_ is None: lambda_ random.uniform(0.3, 0.7) # 3. 图像加权融合核心img1*λ img2*(1-λ) img1_float img1.astype(np.float32) img2_float img2.astype(np.float32) img_mix lambda_ * img1_float (1 - lambda_) * img2_float img_mix np.clip(img_mix, 0, 255).astype(np.uint8) # 4. 标签合并检测任务保留两张图的所有框无需加权 # 分类任务需加权类别概率检测任务直接合并框即可 bboxes_mix np.vstack([bboxes1, bboxes2]) if len(bboxes1) 0 and len(bboxes2) 0 else ( bboxes1 if len(bboxes1) 0 else bboxes2) return img_mix, bboxes_mix # 新增可视化MixUp结果带Box def visualize_mix_up(img_mix: np.ndarray, bboxes_mix: np.ndarray, class_ids: list, save_path: str): 将合并后的YOLO框绘制到MixUp图像上保存带框的可视化结果 :param img_mix: MixUp融合后的BGR图像 :param bboxes_mix: 合并后的YOLO归一化框 [x,y,w,h] :param class_ids: 框对应的类别ID列表 :param save_path: 可视化结果保存路径 img_vis img_mix.copy() h, w img_vis.shape[:2] # 定义不同类别的框颜色区分img1和img2的框 class_colors {0: (0, 255, 0), 1: (0, 0, 255)} # class0绿色class1红色 # 遍历所有框绘制到图像上 for cls, box in zip(class_ids, bboxes_mix): # YOLO归一化框 → 像素坐标 x_center, y_center, box_w, box_h box x_center_pix x_center * w y_center_pix y_center * h box_w_pix box_w * w box_h_pix box_h * h # 计算框的左上角/右下角坐标 x1 int(x_center_pix - box_w_pix / 2) y1 int(y_center_pix - box_h_pix / 2) x2 int(x_center_pix box_w_pix / 2) y2 int(y_center_pix box_h_pix / 2) # 绘制矩形框 类别标签 color class_colors.get(cls, (255, 0, 0)) # 默认蓝色 cv2.rectangle(img_vis, (x1, y1), (x2, y2), color, 2) cv2.putText(img_vis, fcls{cls}, (x1, y1 - 5), cv2.FONT_HERSHEY_SIMPLEX, 0.5, color, 2) # 保存可视化结果 cv2.imwrite(save_path, img_vis) print(f带Box的MixUp可视化结果已保存{save_path}) # 极简加载/保存YOLO标签 def load_yolo_bboxes(label_path: str) - np.ndarray: 仅加载YOLO归一化框 bboxes [] if os.path.exists(label_path): with open(label_path) as f: for line in f: parts line.strip().split() if len(parts) 5: bboxes.append([float(x) for x in parts[1:5]]) return np.array(bboxes, dtypenp.float32) def save_yolo_bboxes(label_path: str, class_ids: list, bboxes: np.ndarray): 保存MixUp后的YOLO标签合并两张图的类别框 # class_ids需和bboxes长度匹配如img1的框对应class0img2的框对应class1 with open(label_path, w) as f: for cls, box in zip(class_ids, bboxes): f.write(f{cls} {box[0]:.6f} {box[1]:.6f} {box[2]:.6f} {box[3]:.6f}\n) # 核心调用示例 if __name__ __main__: # 1. 配置两张图/标签路径MixUp需至少两张图 IMG1_PATH rC:\Users\Excub\workspace\Pytorch\12.08_cv\jjy.jpg LABEL1_PATH rC:\Users\Excub\workspace\Pytorch\12.08_cv\out\jjy\jjy.txt # class 0 IMG2_PATH rC:\Users\Excub\workspace\Pytorch\12.08_cv\jjy1.jpg LABEL2_PATH rC:\Users\Excub\workspace\Pytorch\12.08_cv\out\jjy\jjy1.txt # class 1 SAVE_DIR rC:\Users\Excub\workspace\Pytorch\12.08_cv\out # 2. 加载两张图标签 img1 cv2.imread(IMG1_PATH) img2 cv2.imread(IMG2_PATH) bboxes1 load_yolo_bboxes(LABEL1_PATH) bboxes2 load_yolo_bboxes(LABEL2_PATH) # 3. 核心调用MixUp融合 img_mix, bboxes_mix mix_up_images(img1, img2, bboxes1, bboxes2, lambda_0.5) # 4. 准备合并后的类别IDimg1的框→class0img2的框→class1 class_ids [0] * len(bboxes1) [1] * len(bboxes2) # 5. 保存结果原融合图 带Box的可视化图 os.makedirs(SAVE_DIR, exist_okTrue) # 保存纯融合图 mix_img_path f{SAVE_DIR}/mix_up_result.jpg cv2.imwrite(mix_img_path, img_mix) # 保存带Box的可视化图 vis_img_path f{SAVE_DIR}/mix_up_result_with_box.jpg visualize_mix_up(img_mix, bboxes_mix, class_ids, vis_img_path) # 6. 保存标签 save_yolo_bboxes(f{SAVE_DIR}/mix_up_result.txt, class_ids, bboxes_mix) print(f\nMixUp完成融合系数λ0.5合并后框数{len(bboxes_mix)}) print(生成文件列表) print(f1. 纯融合图像{mix_img_path}) print(f2. 带Box可视化图像{vis_img_path}) print(f3. 合并后标签文件{SAVE_DIR}/mix_up_result.txt) print(\n核心要点) print(1. 核心作用提升模型泛化性缓解过拟合尤其小样本工业场景) print(2. 关键约束两张图需同尺寸自动resize适配λ取0.3~0.7效果最优) print(3. 标签处理检测任务直接合并框分类任务需加权类别概率) print(4. 可视化说明class0img1框为绿色class1img2框为红色便于区分来源) print(5. 区别于普通融合MixUp是随机λ加权且标签需同步融合非简单拼接。)2.mosaic数据增强本质将四张图组合到一起形成新的图遵循“随机采样策略约束”裁剪面积要求保证有效的box留存率。本质就是在数据集中随机选4张图片再组合起来形成新的图片增加box的密度。#!/usr/bin/env python # -*- coding: utf-8 -*- Project Pytorch File mosaic.py IDE PyCharm Author wjj Date 2025/12/14 22:00 Description:Mosaic增强4图固定位置保留所有Box修复NameError import cv2 import os import numpy as np # 单张图预处理resizeBox坐标转换 def process_single_image(img_path, label_path, target_region_size): 处理单张图 1. resize到目标区域尺寸最终图的1/4 2. 转换Box坐标为区域内像素坐标 3. 返回处理后的图、区域内Box像素坐标、类别ID region_w, region_h target_region_size # 加载图像并resize到区域尺寸 img cv2.imread(img_path) if img is None: img np.ones((region_h, region_w, 3), dtypenp.uint8) * 128 # 灰色填充 img cv2.resize(img, (region_w, region_h)) # 加载标签并转换为区域内像素坐标 box_pix [] # 区域内像素坐标[x1, y1, x2, y2] class_ids [] if os.path.exists(label_path): with open(label_path, r) as f: for line in f: line line.strip() if not line: continue parts line.split() try: if len(parts) 5: cls_id int(parts[0]) # 归一化坐标→区域内像素坐标 xc, yc, bw, bh [float(p) for p in parts[1:5]] x1 int((xc - bw / 2) * region_w) y1 int((yc - bh / 2) * region_h) x2 int((xc bw / 2) * region_w) y2 int((yc bh / 2) * region_h) # 确保Box在区域内 x1 max(0, x1) y1 max(0, y1) x2 min(region_w - 1, x2) y2 min(region_h - 1, y2) box_pix.append([x1, y1, x2, y2]) class_ids.append(cls_id) except (ValueError, IndexError): print(f警告标签文件{label_path}无效行{line}) continue return img, box_pix, class_ids # 核心Mosaic拼接固定4图位置保留所有Box def fixed_position_mosaic(img_paths, label_paths, target_size(640, 640)): 固定位置Mosaic - 图0 → 左上区域0~w/2, 0~h/2 - 图1 → 右上区域w/2~w, 0~h/2 - 图2 → 左下区域0~w/2, h/2~h - 图3 → 右下区域w/2~w, h/2~h 强制保留所有图的Box转换为最终图的归一化坐标 assert len(img_paths) 4 and len(label_paths) 4, 必须传入4张图/标签 final_w, final_h target_size region_w final_w // 2 region_h final_h // 2 # 步骤1创建最终Mosaic大图 mosaic_img np.zeros((final_h, final_w, 3), dtypenp.uint8) # 步骤2处理每张图并拼接同时转换Box坐标到最终图 all_final_boxes [] # 最终图的归一化Box all_class_ids [] # 最终图的类别ID box_counts_per_img [0, 0, 0, 0] # 记录每张图的Box数量 # 图0 → 左上区域 img0, boxes0, cls0 process_single_image(img_paths[0], label_paths[0], (region_w, region_h)) mosaic_img[0:region_h, 0:region_w] img0 box_counts_per_img[0] len(boxes0) # 转换Box到最终图坐标 for (x1, y1, x2, y2), cls in zip(boxes0, cls0): xc (x1 x2) / 2 / final_w yc (y1 y2) / 2 / final_h bw (x2 - x1) / final_w bh (y2 - y1) / final_h all_final_boxes.append([xc, yc, bw, bh]) all_class_ids.append(cls) # 图1 → 右上区域 img1, boxes1, cls1 process_single_image(img_paths[1], label_paths[1], (region_w, region_h)) mosaic_img[0:region_h, region_w:final_w] img1 box_counts_per_img[1] len(boxes1) # 转换Box到最终图坐标x坐标region_w for (x1, y1, x2, y2), cls in zip(boxes1, cls1): x1_final x1 region_w x2_final x2 region_w xc (x1_final x2_final) / 2 / final_w yc (y1 y2) / 2 / final_h bw (x2_final - x1_final) / final_w bh (y2 - y1) / final_h all_final_boxes.append([xc, yc, bw, bh]) all_class_ids.append(cls) # 图2 → 左下区域 img2, boxes2, cls2 process_single_image(img_paths[2], label_paths[2], (region_w, region_h)) mosaic_img[region_h:final_h, 0:region_w] img2 box_counts_per_img[2] len(boxes2) # 转换Box到最终图坐标y坐标region_h for (x1, y1, x2, y2), cls in zip(boxes2, cls2): y1_final y1 region_h y2_final y2 region_h xc (x1 x2) / 2 / final_w yc (y1_final y2_final) / 2 / final_h bw (x2 - x1) / final_w bh (y2_final - y1_final) / final_h all_final_boxes.append([xc, yc, bw, bh]) all_class_ids.append(cls) # 图3 → 右下区域 img3, boxes3, cls3 process_single_image(img_paths[3], label_paths[3], (region_w, region_h)) mosaic_img[region_h:final_h, region_w:final_w] img3 box_counts_per_img[3] len(boxes3) # 转换Box到最终图坐标xregion_w, yregion_h for (x1, y1, x2, y2), cls in zip(boxes3, cls3): x1_final x1 region_w x2_final x2 region_w y1_final y1 region_h y2_final y2 region_h xc (x1_final x2_final) / 2 / final_w yc (y1_final y2_final) / 2 / final_h bw (x2_final - x1_final) / final_w bh (y2_final - y1_final) / final_h all_final_boxes.append([xc, yc, bw, bh]) all_class_ids.append(cls) # 转换为numpy数组 all_final_boxes np.array(all_final_boxes, dtypenp.float32) return mosaic_img, all_final_boxes, all_class_ids, box_counts_per_img # 可视化绘制所有Box def visualize_mosaic_with_all_boxes(mosaic_img, boxes, class_ids, box_counts_per_img, save_path): 修复NameError正确统计每张图的Box数量 :param mosaic_img: 最终拼接图 :param boxes: 所有Box的归一化坐标 :param class_ids: 所有Box的类别ID :param box_counts_per_img: 每张图的Box数量列表 [img0, img1, img2, img3] :param save_path: 可视化图保存路径 img_vis mosaic_img.copy() final_h, final_w img_vis.shape[:2] # 不同图的Box用不同颜色区分 colors [ (0, 255, 0), # 图0左上→ 绿色 (0, 0, 255), # 图1右上→ 红色 (255, 0, 0), # 图2左下→ 蓝色 (255, 255, 0) # 图3右下→ 黄色 ] # 按图的索引遍历Box box_idx 0 for img_idx in range(4): color colors[img_idx] # 取当前图的Box数量 curr_box_num box_counts_per_img[img_idx] if curr_box_num 0: continue # 遍历当前图的所有Box for _ in range(curr_box_num): if box_idx len(boxes): break box boxes[box_idx] cls class_ids[box_idx] # 转换为像素坐标 xc, yc, bw, bh box x1 int((xc - bw / 2) * final_w) y1 int((yc - bh / 2) * final_h) x2 int((xc bw / 2) * final_w) y2 int((yc bh / 2) * final_h) # 绘制Box和标注 cv2.rectangle(img_vis, (x1, y1), (x2, y2), color, 2) cv2.putText(img_vis, fimg{img_idx}_cls{cls}, (x1, y1 - 5), cv2.FONT_HERSHEY_SIMPLEX, 0.5, color, 2) box_idx 1 # 保存可视化图 cv2.imwrite(save_path, img_vis) print(f带所有Box的Mosaic图已保存{save_path}) print( f各图Box数量img0{box_counts_per_img[0]}, img1{box_counts_per_img[1]}, img2{box_counts_per_img[2]}, img3{box_counts_per_img[3]}) # 保存YOLO标签 def save_yolo_labels(label_path, class_ids, boxes): with open(label_path, w) as f: for cls, box in zip(class_ids, boxes): xc, yc, bw, bh box f.write(f{cls} {xc:.6f} {yc:.6f} {bw:.6f} {bh:.6f}\n) print(f包含所有Box的YOLO标签已保存{label_path}) # 主函数 if __name__ __main__: # 1. 配置4张图/标签路径替换为你的实际路径 IMG_PATHS [ rC:\Users\Excub\workspace\Pytorch\12.08_cv\jjy.jpg, rC:\Users\Excub\workspace\Pytorch\12.08_cv\jjy1.jpg, rC:\Users\Excub\workspace\Pytorch\12.08_cv\jjy2.jpg, rC:\Users\Excub\workspace\Pytorch\12.08_cv\jjy3.jpg, ] LABEL_PATHS [ rC:\Users\Excub\workspace\Pytorch\12.08_cv\out\jjy\jjy.txt, rC:\Users\Excub\workspace\Pytorch\12.08_cv\out\jjy\jjy1.txt, rC:\Users\Excub\workspace\Pytorch\12.08_cv\out\jjy\jjy2.txt, rC:\Users\Excub\workspace\Pytorch\12.08_cv\out\jjy\jjy3.txt ] SAVE_DIR rC:\Users\Excub\workspace\Pytorch\12.08_cv\out FINAL_SIZE (640, 640) # 最终Mosaic图尺寸 # 2. 创建保存目录 os.makedirs(SAVE_DIR, exist_okTrue) # 3. 执行固定位置Mosaic强制保留所有图和Box mosaic_img, all_boxes, all_classes, box_counts fixed_position_mosaic(IMG_PATHS, LABEL_PATHS, FINAL_SIZE) # 4. 保存纯Mosaic图 mosaic_img_path os.path.join(SAVE_DIR, mosaic_fixed_position1.jpg) cv2.imwrite(mosaic_img_path, mosaic_img) print(f纯Mosaic图已保存{mosaic_img_path}) # 5. 保存带所有Box的可视化图修复参数传递 vis_img_path os.path.join(SAVE_DIR, mosaic_with_all_boxes1.jpg) visualize_mosaic_with_all_boxes(mosaic_img, all_boxes, all_classes, box_counts, vis_img_path) # 6. 保存包含所有Box的YOLO标签 label_path os.path.join(SAVE_DIR, mosaic_all_boxes1.txt) save_yolo_labels(label_path, all_classes, all_boxes) # 7. 最终统计 print(f\n 固定位置Mosaic增强完成 ) print(f最终图尺寸{FINAL_SIZE}) print(f总Box数量{len(all_boxes)}4张图的Box全部保留) print(f保存路径{SAVE_DIR}) print(f核心特性) print(f1. 图0→左上图1→右上图2→左下图3→右下位置固定) print(f2. 每张图的Box全部转换为最终图坐标无遗漏) print(f3. 可视化图中不同图的Box用不同颜色区分便于验证) print(f4. 标签文件包含所有Box可直接用于YOLO训练。)3.paste_img1box_to_img2数据增强本质把 img1 中某个 Box 对应的像素区域裁剪出来粘贴到 img2 的指定位置同时将该 Box 的坐标从 img1 的归一化坐标转换为 img2 的归一化坐标生成新标签。步骤核心操作公式 / 逻辑1. Box 解析归一化→像素坐标x1 (xc - bw/2)*w1,y1 (yc - bh/2)*h12. 裁剪区域从 img1 裁剪 Box 区域crop_img img1[y1:y2, x1:x2]3. 粘贴位置避免越界paste_x ∈ [0, w2-crop_w],paste_y ∈ [0, h2-crop_h]4. 图像粘贴覆盖 img2 对应区域img2[paste_y:paste_ycrop_h, paste_x:paste_xcrop_w] crop_img5. 坐标转换像素→归一化img2new_xc (paste_x paste_xcrop_w)/2 /w2import cv2 import numpy as np def paste_img1box_to_img2( img1_path: str, # 源图要裁剪Box的图路径 img2_path: str, # 目标图要粘贴的图路径 label1_path: str, # 源图标签路径包含Box信息 box_idx: int 0, # 选择源图的第几个Box默认第0个 paste_pos: tuple None, # 粘贴位置 (x, y)None则随机位置 save_path: str paste_result.jpg, # 结果保存路径 label_save_path: str paste_result.txt # 新标签保存路径 ): 将img1的指定Box区域粘贴到img2并生成新的标签Box坐标转换为img2的归一化坐标 :return: 粘贴后的图像、新的归一化Box坐标、新标签 # 步骤1读取并预处理图像 # 读取源图img1和目标图img2 img1 cv2.imread(img1_path) img2 cv2.imread(img2_path) if img1 is None: raise ValueError(f源图{img1_path}读取失败) if img2 is None: raise ValueError(f目标图{img2_path}读取失败) h1, w1 img1.shape[:2] # 源图尺寸 h2, w2 img2.shape[:2] # 目标图尺寸 # 步骤2解析img1的Box归一化→像素坐标 box1_norm None # 源图Box的归一化坐标 [xc, yc, bw, bh] box1_pix None # 源图Box的像素坐标 [x1, y1, x2, y2] cls1 None # 源图Box的类别ID with open(label1_path, r) as f: lines [line.strip() for line in f if line.strip()] if box_idx len(lines): raise IndexError(f源图标签仅{len(lines)}个Box无法选择第{box_idx}个) # 解析指定Box parts lines[box_idx].split() if len(parts) 5: raise ValueError(f源图标签行格式错误{lines[box_idx]}) cls1 int(parts[0]) xc, yc, bw, bh [float(p) for p in parts[1:5]] # 归一化坐标→像素坐标x1,y1左上x2,y2右下 x1 int((xc - bw/2) * w1) y1 int((yc - bh/2) * h1) x2 int((xc bw/2) * w1) y2 int((yc bh/2) * h1) # 确保Box在img1范围内 x1 max(0, x1) y1 max(0, y1) x2 min(w1-1, x2) y2 min(h1-1, y2) box1_norm [xc, yc, bw, bh] box1_pix [x1, y1, x2, y2] # 步骤3裁剪img1的Box区域 crop_img img1[y1:y2, x1:x2] # 裁剪Box对应的像素区域 crop_h, crop_w crop_img.shape[:2] if crop_h 0 or crop_w 0: raise ValueError(f源图Box裁剪区域为空Box像素坐标{box1_pix}) # 步骤4确定img2上的粘贴位置避免越界 if paste_pos is None: # 随机位置确保裁剪区域完全在img2内 paste_x np.random.randint(0, w2 - crop_w) paste_y np.random.randint(0, h2 - crop_h) else: paste_x, paste_y paste_pos # 强制修正越界位置 paste_x max(0, min(paste_x, w2 - crop_w)) paste_y max(0, min(paste_y, h2 - crop_h)) # 步骤5将裁剪区域粘贴到img2 img2_pasted img2.copy() img2_pasted[paste_y:paste_ycrop_h, paste_x:paste_xcrop_w] crop_img # 步骤6转换Box坐标为img2的归一化坐标 # 粘贴后的Box像素坐标img2上 new_x1 paste_x new_y1 paste_y new_x2 paste_x crop_w new_y2 paste_y crop_h # 像素坐标→归一化坐标 new_xc (new_x1 new_x2) / 2 / w2 new_yc (new_y1 new_y2) / 2 / h2 new_bw (new_x2 - new_x1) / w2 new_bh (new_y2 - new_y1) / h2 new_box_norm [new_xc, new_yc, new_bw, new_bh] # 步骤7保存结果 # 保存粘贴后的图像 cv2.imwrite(save_path, img2_pasted) print(f粘贴后的图像已保存{save_path}) # 保存新标签格式cls xc yc bw bh with open(label_save_path, w) as f: f.write(f{cls1} {new_xc:.6f} {new_yc:.6f} {new_bw:.6f} {new_bh:.6f}\n) print(f新标签已保存{label_save_path}) return img2_pasted, new_box_norm, cls1 # ------------------- 调用示例 ------------------- if __name__ __main__: # 配置路径 IMG1_PATH rC:\Users\Excub\workspace\Pytorch\12.08_cv\jjy2.jpg # 源图要裁剪的图 IMG2_PATH rC:\Users\Excub\workspace\Pytorch\12.08_cv\jjy1.jpg # 目标图要粘贴的背景图 LABEL1_PATH rC:\Users\Excub\workspace\Pytorch\12.08_cv\out\jjy\jjy2.txt # 源图标签 SAVE_PATH rC:\Users\Excub\workspace\Pytorch\12.08_cv\out\paste_result.jpg LABEL_SAVE_PATH rC:\Users\Excub\workspace\Pytorch\12.08_cv\out\paste_label.txt # 执行粘贴操作将img1的第0个Box粘贴到img2的(100, 100)位置 try: pasted_img, new_box, cls paste_img1box_to_img2( img1_pathIMG1_PATH, img2_pathIMG2_PATH, label1_pathLABEL1_PATH, box_idx0, # 选择img1的第0个Box paste_pos(100, 100), # 粘贴到img2的(100,100)位置 save_pathSAVE_PATH, label_save_pathLABEL_SAVE_PATH ) print(f粘贴完成) print(f新Box归一化坐标xc{new_box[0]:.6f}, yc{new_box[1]:.6f}, bw{new_box[2]:.6f}, bh{new_box[3]:.6f}) print(fBox类别{cls}) except Exception as e: print(f执行失败{e})3.局部增强/裁剪类数据增强锦集1.sliding_crop本质从一张大图中按固定步长滑动裁剪出多个子图同时将Box坐标转换到对应的子图中是目标检测中扩充数据/处理大图的常用手段。「按固定窗口 步长遍历大图→裁剪子图→转换原图 Box 到子图坐标」#!/usr/bin/env python # -*- coding: utf-8 -*- Project Pytorch File sliding_crop.py IDE PyCharm Author wjj Date 2025/12/15 16:43 Description: Sliding Crop (Fix encoding window calculation) import cv2 import numpy as np import os from typing import List, Tuple, Dict def sliding_crop( img_path: str, label_path: str, crop_size: Tuple[int, int] (640, 640), # Crop window size (w, h) stride: Tuple[int, int] (320, 320), # Sliding stride (x, y) save_dir: str sliding_crop_album, # Album save dir keep_partial_box: bool True, # Keep box if center in window filter_empty: bool True, # Filter sub-img without box class_names: List[str] None # Class names (optional) ) - List[Dict]: Sliding crop large image to sub-images, convert box coords to sub-img normalized coords :return: Crop results list (sub-img path, label path, box num, etc.) # Initialization os.makedirs(save_dir, exist_okTrue) crop_w, crop_h crop_size stride_x, stride_y stride result_list [] # Read image label # Read large image img cv2.imread(img_path) if img is None: raise ValueError(fFailed to read image: {img_path}!) img_h, img_w img.shape[:2] # Auto adjust oversize crop window if crop_w img_w or crop_h img_h: print( fWarning: Crop window ({crop_w}x{crop_h}) ≥ image size ({img_w}x{img_h}), auto adjust to 1/2 of image size) crop_w img_w // 2 crop_h img_h // 2 stride_x crop_w // 2 stride_y crop_h // 2 print(fAdjusted: Crop window ({crop_w}x{crop_h}), stride ({stride_x}x{stride_y})) # Parse label (normalized → pixel coords, GBK encoding for Windows) original_boxes [] try: with open(label_path, r, encodinggbk) as f: for line in f: line line.strip() if not line: continue parts line.split() if len(parts) 5: print(fWarning: Invalid label line {line}, skip) continue try: cls_id int(parts[0]) xc, yc, bw, bh [float(p) for p in parts[1:5]] # Normalized → pixel coords x1 int((xc - bw / 2) * img_w) y1 int((yc - bh / 2) * img_h) x2 int((xc bw / 2) * img_w) y2 int((yc bh / 2) * img_h) # Fix out-of-bounds x1 max(0, x1) y1 max(0, y1) x2 min(img_w - 1, x2) y2 min(img_h - 1, y2) original_boxes.append((x1, y1, x2, y2, cls_id)) except Exception as e: print(fFailed to parse label line {line}: {e}) continue except FileNotFoundError: raise FileNotFoundError(fLabel file {label_path} not found!) # Calculate sliding window positions # X direction starts x_starts [] x 0 while x crop_w img_w: x_starts.append(x) x stride_x if x_starts and x_starts[-1] crop_w img_w: x_starts.append(img_w - crop_w) if not x_starts: x_starts [0] # Y direction starts y_starts [] y 0 while y crop_h img_h: y_starts.append(y) y stride_y if y_starts and y_starts[-1] crop_h img_h: y_starts.append(img_h - crop_h) if not y_starts: y_starts [0] # Crop sub-images crop_idx 0 for y_start in y_starts: for x_start in x_starts: x_end x_start crop_w y_end y_start crop_h # Crop sub-image sub_img img[y_start:y_end, x_start:x_end] sub_img_h, sub_img_w sub_img.shape[:2] if sub_img_h ! crop_h or sub_img_w ! crop_w: # Pad black border to keep size sub_img cv2.copyMakeBorder( sub_img, 0, crop_h - sub_img_h, 0, crop_w - sub_img_w, cv2.BORDER_CONSTANT, value(0, 0, 0) ) # Filter boxes in current window sub_boxes [] for (bx1, by1, bx2, by2, cls_id) in original_boxes: box_cx (bx1 bx2) / 2 box_cy (by1 by2) / 2 if keep_partial_box: in_window (x_start box_cx x_end) and (y_start box_cy y_end) else: in_window (x_start bx1) and (bx2 x_end) and (y_start by1) and (by2 y_end) if not in_window: continue # Convert to sub-image normalized coords sub_bx1 bx1 - x_start sub_by1 by1 - y_start sub_bx2 bx2 - x_start sub_by2 by2 - y_start # Fix out-of-bounds in sub-image sub_bx1 max(0, sub_bx1) sub_by1 max(0, sub_by1) sub_bx2 min(crop_w - 1, sub_bx2) sub_by2 min(crop_h - 1, sub_by2) # Pixel → normalized sub_xc (sub_bx1 sub_bx2) / 2 / crop_w sub_yc (sub_by1 sub_by2) / 2 / crop_h sub_bw (sub_bx2 - sub_bx1) / crop_w sub_bh (sub_by2 - sub_by1) / crop_h sub_boxes.append((sub_xc, sub_yc, sub_bw, sub_bh, cls_id)) # Filter empty sub-images if filter_empty and len(sub_boxes) 0: continue # Save sub-image and label sub_img_name fcrop_{crop_idx:04d}_x{x_start}_y{y_start}.jpg sub_label_name fcrop_{crop_idx:04d}_x{x_start}_y{y_start}.txt sub_img_path os.path.join(save_dir, sub_img_name) sub_label_path os.path.join(save_dir, sub_label_name) cv2.imwrite(sub_img_path, sub_img) # Save label with GBK encoding with open(sub_label_path, w, encodinggbk) as f: for (xc, yc, bw, bh, cls_id) in sub_boxes: f.write(f{cls_id} {xc:.6f} {yc:.6f} {bw:.6f} {bh:.6f}\n) # Record result result_info { crop_idx: crop_idx, window_pos: (x_start, y_start, x_end, y_end), sub_img_path: sub_img_path, sub_label_path: sub_label_path, box_num: len(sub_boxes) } result_list.append(result_info) print(fCrop completed {sub_img_path} | Box count: {len(sub_boxes)}) crop_idx 1 # Generate summary file (GBK encoding) album_summary_path os.path.join(save_dir, sliding_crop_summary.txt) with open(album_summary_path, w, encodinggbk) as f: f.write(Sliding Crop Album Summary\n) f.write(fOriginal image path: {img_path}\n) f.write(fOriginal image size: {img_w}x{img_h}\n) f.write(fCrop window size: {crop_w}x{crop_h}\n) f.write(fSliding stride: {stride_x}x{stride_y}\n) f.write(fGenerated sub-images: {len(result_list)}\n) f.write(- * 50 \n) for info in result_list: f.write( fIndex {info[crop_idx]} | Window pos {info[window_pos]} | Sub-img {info[sub_img_path]} | Box count {info[box_num]}\n) print(f\nSliding crop completed!) print(fAlbum save directory: {save_dir}) print(fGenerated sub-images count: {len(result_list)}) print(fSummary file: {album_summary_path}) return result_list # ------------------- Test Example ------------------- if __name__ __main__: # Config IMG_PATH rC:\Users\Excub\workspace\Pytorch\12.08_cv\jjy1.jpg LABEL_PATH rC:\Users\Excub\workspace\Pytorch\12.08_cv\out\jjy\jjy1.txt SAVE_DIR rC:\Users\Excub\workspace\Pytorch\12.08_cv\out\sliding_crop_album CROP_SIZE (640, 640) # Reasonable size for model input STRIDE (320, 320) # 50% overlap # Run sliding crop try: crop_results sliding_crop( img_pathIMG_PATH, label_pathLABEL_PATH, crop_sizeCROP_SIZE, strideSTRIDE, save_dirSAVE_DIR, keep_partial_boxTrue, filter_emptyTrue ) except Exception as e: print(fSliding crop failed: {e})实现效果1-4的核心对比裁剪后的子图能实现数据增强样本数量扩充从 1 张原图→4 张子图直接增加训练样本量降低模型过拟合风险目标位置多样性子图中目标的「相对位置」和原图不同比如原图中目标在中间子图中目标在左上角 / 右下角模型能学习到「目标出现在图像任意位置」的特征避免对目标位置过拟合目标聚焦增强如果原图中目标占比小比如小目标裁剪后的子图会让目标在画面中占比更高模型更容易捕捉目标的细节特征抗截断鲁棒性50% 重叠的裁剪方式会让跨窗口的目标在不同子图中被「部分保留」比如目标一半在 A 子图、一半在 B 子图模型能学习到「目标被截断时的特征」提升检测的鲁棒性。2.copy_paste本质在同一张图内复制指定类别目标随机粘贴到空白区域同图内复制指定类别目标→随机粘贴到空白区域支持翻转 / 旋转 / 缩放增强且通过掩码避免目标重叠。#!/usr/bin/env python # -*- coding: utf-8 -*- Project Pytorch File copy_paste.py IDE PyCharm Author wjj Date 2025/12/15 17:12 Description: import cv2 import numpy as np import os import random from typing import List, Tuple def copy_paste( img_path: str, label_path: str, target_cls: int, # 要复制的目标类别ID如0 paste_num: int 3, # 复制粘贴的数量默认3个 save_path: str copy_paste_aug.jpg, new_label_path: str copy_paste_aug.txt, aug_prob: float 0.5, # 目标增强翻转/旋转/缩放的概率 scale_range: Tuple (0.8, 1.2), # 缩放范围 rotate_range: Tuple (-15, 15) # 旋转角度范围±15° ): 同图内复制指定类别目标随机粘贴到空白区域避免重叠支持目标增强 :param target_cls: 要增强的稀有目标类别ID :param paste_num: 复制粘贴的目标数量 # 1. 读取图片和标签 # 读取原图 img cv2.imread(img_path) if img is None: raise ValueError(f读取图片失败: {img_path}) img_h, img_w img.shape[:2] img_copy img.copy() # 用于粘贴的画布 # 解析标签分离目标类别 转换为像素坐标 original_boxes [] # 所有目标[(x1,y1,x2,y2,cls_id), ...] target_boxes [] # 指定类别的目标[(x1,y1,x2,y2), ...] with open(label_path, r, encodinggbk) as f: for line in f: parts line.strip().split() if len(parts) 5: continue cls_id int(parts[0]) xc, yc, bw, bh [float(p) for p in parts[1:5]] # 归一化→像素坐标 x1 int((xc - bw / 2) * img_w) y1 int((yc - bh / 2) * img_h) x2 int((xc bw / 2) * img_w) y2 int((yc bh / 2) * img_h) original_boxes.append((x1, y1, x2, y2, cls_id)) # 筛选指定类别的目标 if cls_id target_cls: target_boxes.append((x1, y1, x2, y2)) # 校验指定类别是否有目标 if len(target_boxes) 0: raise ValueError(f图片中无类别ID为 {target_cls} 的目标) # 2. 生成空白区域掩码避免粘贴重叠 # 初始化掩码0空白1已有目标 mask np.zeros((img_h, img_w), dtypenp.uint8) for (x1, y1, x2, y2, _) in original_boxes: mask[y1:y2, x1:x2] 1 # 已有目标区域标记为1 # 3. 复制指定类别目标并增强 # 随机选一个指定类别目标作为复制模板若有多个随机选 src_x1, src_y1, src_x2, src_y2 random.choice(target_boxes) src_crop img[src_y1:src_y2, src_x1:src_x2] # 复制的目标像素区域 src_h, src_w src_crop.shape[:2] # 存储粘贴后的新目标坐标用于生成新标签 new_boxes [] # 4. 随机粘贴到空白区域 paste_count 0 max_attempts paste_num * 10 # 最大尝试次数避免死循环 attempts 0 while paste_count paste_num and attempts max_attempts: attempts 1 # -------- 4.1 对复制的目标做增强翻转/旋转/缩放 -------- aug_crop src_crop.copy() # 随机水平翻转 if random.random() aug_prob: aug_crop cv2.flip(aug_crop, 1) # 随机缩放 scale random.uniform(*scale_range) aug_w int(src_w * scale) aug_h int(src_h * scale) aug_crop cv2.resize(aug_crop, (aug_w, aug_h)) # 随机旋转带黑边避免裁剪 if random.random() aug_prob: angle random.uniform(*rotate_range) M cv2.getRotationMatrix2D((aug_w / 2, aug_h / 2), angle, 1) aug_crop cv2.warpAffine(aug_crop, M, (aug_w, aug_h), borderValue(0, 0, 0)) # -------- 4.2 随机选择空白粘贴位置 -------- # 确保粘贴位置在图片内且目标完整放下 paste_x random.randint(0, img_w - aug_w) paste_y random.randint(0, img_h - aug_h) # 检查粘贴区域是否为空白掩码全0 paste_mask mask[paste_y:paste_y aug_h, paste_x:paste_x aug_w] if np.sum(paste_mask) 0: # 无重叠可粘贴 # 粘贴目标到画布 img_copy[paste_y:paste_y aug_h, paste_x:paste_x aug_w] aug_crop # 更新掩码标记为已有目标 mask[paste_y:paste_y aug_h, paste_x:paste_x aug_w] 1 # 记录新目标坐标像素→归一化 new_xc (paste_x aug_w / 2) / img_w new_yc (paste_y aug_h / 2) / img_h new_bw aug_w / img_w new_bh aug_h / img_h new_boxes.append((target_cls, new_xc, new_yc, new_bw, new_bh)) paste_count 1 print(f成功粘贴第 {paste_count} 个目标 → 位置: ({paste_x}, {paste_y})) if paste_count paste_num: print(f警告仅成功粘贴 {paste_count} 个目标剩余位置有重叠) # 5. 生成新标签原目标 新粘贴目标 with open(new_label_path, w, encodinggbk) as f: # 写入原目标 for (x1, y1, x2, y2, cls_id) in original_boxes: xc (x1 x2) / 2 / img_w yc (y1 y2) / 2 / img_h bw (x2 - x1) / img_w bh (y2 - y1) / img_h f.write(f{cls_id} {xc:.6f} {yc:.6f} {bw:.6f} {bh:.6f}\n) # 写入新粘贴的目标 for (cls_id, xc, yc, bw, bh) in new_boxes: f.write(f{cls_id} {xc:.6f} {yc:.6f} {bw:.6f} {bh:.6f}\n) # 6. 保存增强后的图片 cv2.imwrite(save_path, img_copy) print(f\nCopy-Paste 增强完成) print(f增强后图片保存至: {save_path}) print(f新标签保存至: {new_label_path}) print(f原目标数量: {len(original_boxes)} | 新增目标数量: {len(new_boxes)}) return img_copy, new_boxes # ------------------- 测试示例 ------------------- if __name__ __main__: # 配置参数仅需修改以下5项 IMG_PATH rC:\Users\Excub\workspace\Pytorch\12.08_cv\jjy1.jpg LABEL_PATH rC:\Users\Excub\workspace\Pytorch\12.08_cv\out\jjy\jjy1.txt TARGET_CLS 0 # 要增强的目标类别ID根据你的标签修改 SAVE_PATH rC:\Users\Excub\workspace\Pytorch\12.08_cv\out\copy_paste_aug.jpg NEW_LABEL_PATH rC:\Users\Excub\workspace\Pytorch\12.08_cv\out\copy_paste_aug.txt # 核心调用 try: copy_paste( img_pathIMG_PATH, label_pathLABEL_PATH, target_clsTARGET_CLS, paste_num3, # 复制粘贴3个目标 save_pathSAVE_PATH, new_label_pathNEW_LABEL_PATH ) except Exception as e: print(f执行失败: {e})修改前后对比模块实现逻辑目标筛选解析标签后筛选指定target_cls的目标仅复制该类别解决稀有目标增强重叠避免生成掩码矩阵标记已有目标区域仅在掩码为 0 的空白区域粘贴目标增强复制的目标随机做水平翻转50% 概率、缩放0.8~1.2 倍、小角度旋转坐标转换粘贴后自动将像素坐标转换为 YOLO 格式的归一化坐标生成新标签鲁棒性限制最大尝试次数避免因空白区域不足导致死循环同时提示实际粘贴数量3.pyr_up/pyr_down本质很简单就是我们常见的将原图放大或缩小达到增加数据样本的作用。上采样小目标增强建议仅用 1~2 次1 次缩放后目标尺寸 ×2小目标细节更清晰2 次后可能模糊需结合实际图像质量判断适用场景原图中小目标占比 5%检测模型难以捕捉的场景。下采样大目标适配无次数限制建议 1~3 次每次缩小为原来的 1/2适配模型输入尺寸如 640×640适用场景原图中大目标占比 80%超出模型检测范围的场景。验证方法运行代码后打开缩放后的图片对比标注框是否与目标位置匹配检查新标签文件的归一化坐标确保数值在 0~1 范围内无越界import cv2 import numpy as np def pyr_scale( img_path: str, label_path: str, scale_type: str up, # up放大/down缩小 scale_times: int 1, # 缩放次数up建议1次 save_path: str pyr_scaled.jpg, new_label_path: str pyr_scaled.txt ): # 1. 读取原图 img cv2.imread(img_path) orig_h, orig_w img.shape[:2] scaled_img img.copy() # 2. 解析原标签归一化→像素坐标 orig_boxes [] with open(label_path, r, encodinggbk) as f: for line in f: parts line.strip().split() if len(parts) 5: continue cls_id int(parts[0]) xc, yc, bw, bh [float(p) for p in parts[1:5]] x1 int((xc - bw/2) * orig_w) y1 int((yc - bh/2) * orig_h) x2 int((xc bw/2) * orig_w) y2 int((yc bh/2) * orig_h) orig_boxes.append((x1, y1, x2, y2, cls_id)) # 3. 执行图像缩放pyrUp/pyrDown scale_factor 1.0 for _ in range(scale_times): if scale_type up: scaled_img cv2.pyrUp(scaled_img) scale_factor * 2 else: scaled_img cv2.pyrDown(scaled_img) scale_factor / 2 scaled_h, scaled_w scaled_img.shape[:2] # 4. 缩放标注框并生成新标签 with open(new_label_path, w, encodinggbk) as f: for (x1, y1, x2, y2, cls_id) in orig_boxes: # 像素坐标缩放 修正越界 new_x1 max(0, int(x1 * scale_factor)) new_y1 max(0, int(y1 * scale_factor)) new_x2 min(scaled_w-1, int(x2 * scale_factor)) new_y2 min(scaled_h-1, int(y2 * scale_factor)) # 像素→归一化坐标 new_xc (new_x1 new_x2)/2 / scaled_w new_yc (new_y1 new_y2)/2 / scaled_h new_bw (new_x2 - new_x1)/scaled_w new_bh (new_y2 - new_y1)/scaled_h f.write(f{cls_id} {new_xc:.6f} {new_yc:.6f} {new_bw:.6f} {new_bh:.6f}\n) # 5. 保存结果 cv2.imwrite(save_path, scaled_img) print(f完成原图尺寸:{orig_w}x{orig_h} → 缩放后:{scaled_w}x{scaled_h}) print(f缩放后图片{save_path} | 新标签{new_label_path}) # ------------------- 极简验证调用 ------------------- if __name__ __main__: # 仅需修改这6行路径/参数即可验证 IMG_PATH rC:\Users\Excub\workspace\Pytorch\12.08_cv\jjy1.jpg LABEL_PATH rC:\Users\Excub\workspace\Pytorch\12.08_cv\out\jjy\jjy1.txt SCALE_TYPE up # 测试放大改down测试缩小 SCALE_TIMES 1 # 仅缩放1次避免模糊 SAVE_PATH rC:\Users\Excub\workspace\Pytorch\12.08_cv\out\pyr_up.jpg NEW_LABEL_PATH rC:\Users\Excub\workspace\Pytorch\12.08_cv\out\pyr_up.txt pyr_scale( img_pathIMG_PATH, label_pathLABEL_PATH, scale_typeSCALE_TYPE, scale_timesSCALE_TIMES, save_pathSAVE_PATH, new_label_pathNEW_LABEL_PATH )图片与原图一致仅变大/变小罢了。4.cut_box本质「裁剪小目标周边区域 放大」图片变化原图 → 裁剪「目标 周边 100 像素」的局部区域 → 放大 2 倍小目标特征被强化更清晰对比IMG_PATH原图和SAVE_PATH裁剪放大图能直观看到小目标从 “模糊小点” 变成 “清晰区域”。Label 变化原标签目标坐标是「相对于整张原图」的归一化值新标签目标坐标是「相对于裁剪放大后局部图」的归一化值框的占比会显著变大因为图片尺寸变小目标占比提升import cv2 import numpy as np def cut_box( img_path: str, label_path: str, target_cls: int 0, # 要增强的小目标类别ID expand_pixel: int 100, # 目标周边保留像素默认100 scale_factor: int 2, # 放大倍数强化小目标特征 save_path: str cut_box_aug.jpg, new_label_path: str cut_box_aug.txt ): # 1. 读取原图 img cv2.imread(img_path) if img is None: raise ValueError(f读取图片失败: {img_path}) img_h, img_w img.shape[:2] # 2. 解析标签归一化→像素坐标筛选指定类别小目标 target_box None with open(label_path, r, encodinggbk) as f: for line in f: parts line.strip().split() if len(parts) 5: continue cls_id int(parts[0]) if cls_id ! target_cls: continue # 仅处理指定类别 xc, yc, bw, bh [float(p) for p in parts[1:5]] # 归一化→像素坐标 x1 int((xc - bw/2) * img_w) y1 int((yc - bh/2) * img_h) x2 int((xc bw/2) * img_w) y2 int((yc bh/2) * img_h) target_box (x1, y1, x2, y2) break # 仅处理第一个指定类别目标验证用 if target_box is None: raise ValueError(f未找到类别ID为 {target_cls} 的目标) x1, y1, x2, y2 target_box # 3. 裁剪目标周边区域保留expand_pixel像素 crop_x1 max(0, x1 - expand_pixel) crop_y1 max(0, y1 - expand_pixel) crop_x2 min(img_w - 1, x2 expand_pixel) crop_y2 min(img_h - 1, y2 expand_pixel) cut_img img[crop_y1:crop_y2, crop_x1:crop_x2] # 裁剪区域 # 4. 放大裁剪区域强化小目标特征 cut_h, cut_w cut_img.shape[:2] scaled_cut cv2.resize(cut_img, (cut_w * scale_factor, cut_h * scale_factor)) scaled_h, scaled_w scaled_cut.shape[:2] # 5. 重新计算目标框坐标相对于放大后的裁剪图 rel_x1 x1 - crop_x1 rel_y1 y1 - crop_y1 rel_x2 x2 - crop_x1 rel_y2 y2 - crop_y1 # 放大后像素坐标用于绘制框 new_x1 int(rel_x1 * scale_factor) new_y1 int(rel_y1 * scale_factor) new_x2 int(rel_x2 * scale_factor) new_y2 int(rel_y2 * scale_factor) # 归一化坐标保存到标签 new_xc (new_x1 new_x2) / 2 / scaled_w new_yc (new_y1 new_y2) / 2 / scaled_h new_bw (new_x2 - new_x1) / scaled_w new_bh (new_y2 - new_y1) / scaled_h # 新增在放大后的图片上绘制label框 # 绘制红色矩形框线宽2醒目 cv2.rectangle(scaled_cut, (new_x1, new_y1), (new_x2, new_y2), (0, 0, 255), 2) # 可选添加类别文字标注 cv2.putText(scaled_cut, fcls_{target_cls}, (new_x1, new_y1-10), cv2.FONT_HERSHEY_SIMPLEX, 0.6, (0, 0, 255), 2) # 6. 保存结果带框的图片新标签 cv2.imwrite(save_path, scaled_cut) with open(new_label_path, w, encodinggbk) as f: f.write(f{target_cls} {new_xc:.6f} {new_yc:.6f} {new_bw:.6f} {new_bh:.6f}\n) # 输出关键信息对比用 print(f裁剪区域({crop_x1},{crop_y1})→({crop_x2},{crop_y2})) print(f裁剪后尺寸{cut_w}x{cut_h} → 放大后{scaled_w}x{scaled_h}) print(f绘制框坐标({new_x1},{new_y1})→({new_x2},{new_y2})) print(f完成带框图片{save_path} | 新标签{new_label_path}) # ------------------- 极简验证调用 ------------------- if __name__ __main__: # 仅需修改以下6行参数即可验证 IMG_PATH rC:\Users\Excub\workspace\Pytorch\12.08_cv\jjy3.jpg LABEL_PATH rC:\Users\Excub\workspace\Pytorch\12.08_cv\out\jjy\jjy3.txt TARGET_CLS 0 # 要增强的小目标类别ID EXPAND_PIXEL 100 # 目标周边保留100像素 SAVE_PATH rC:\Users\Excub\workspace\Pytorch\12.08_cv\out\cut_box_aug.jpg NEW_LABEL_PATH rC:\Users\Excub\workspace\Pytorch\12.08_cv\out\cut_box_aug.txt cut_box( img_pathIMG_PATH, label_pathLABEL_PATH, target_clsTARGET_CLS, expand_pixelEXPAND_PIXEL, save_pathSAVE_PATH, new_label_pathNEW_LABEL_PATH )成品图片总结本章节承接上一份文章继续总结了常见的数据增强的方法并配上了相应的案例和实例仅供大家学习参考如有不对的地方欢迎大家指出感谢