根据目标检测结果裁剪bbox保存到本地，python，opencv

最新推荐文章于 2025-06-10 15:07:17 发布

TravisPan

最新推荐文章于 2025-06-10 15:07:17 发布

阅读量5.7k

点赞数 2

CC 4.0 BY-SA版权

分类专栏：脚本文章标签：脚本 opencv bbox

本文链接：https://blog.youkuaiyun.com/TravisPan/article/details/88897134

近日有一个需求是把检测的框的图像都截下来保存，就写了个脚本，分享出来

基本逻辑为循环读取每张图片的结果坐标文件，读取框坐标，再用OpenCV的函数裁剪下来并保存。
注意坐标文件为VOC格式，自行百度。

# coding=utf-8
import cv2
import os
import glob
import numpy as np
import re

def CutImg(imgpath,destpath):
	img = glob.glob(r"/DATACENTER6/ph/traffic_model_faster101/in_pic/29-6-ori/*.jpg")
	img_len= int(len(img))
	#get all the images 
	
	for i in range(img_len):
		temp=img[i].rstrip(".jpg")				# get image name to find its' annotation file.
		
		ori_img = cv2.imread(str(img[i]))
		sp = ori_img.shape						#obtain the image shape
		sz1 = sp[0] 							#height(rows) of image
		sz2 = sp[1] 							#width(colums) of image		

		np.set_printoptions(suppress=True)
		txt = np.loadtxt(str(temp)+".txt")		#load the pic's annotation, which is voc-style.
		if txt.shape[0]==0:						#if the annotation is null, drop it.
			continue
		if np.ndim(txt) == 1:					
			txt = txt.reshape(1, 5)

		txt_len=len(txt)
		# print(txt)
		# print(txt_le