00001 import xml.etree.ElementTree as ET 00002 import pickle 00003 import os 00004 from os import listdir, getcwd 00005 from os.path import join 00006 00007 sets=[('2012', 'train'), ('2012', 'val'), ('2007', 'train'), ('2007', 'val'), ('2007', 'test')] 00008 00009 classes = ["aeroplane", "bicycle", "bird", "boat", "bottle", "bus", "car", "cat", "chair", "cow", "diningtable", "dog", "horse", "motorbike", "person", "pottedplant", "sheep", "sofa", "train", "tvmonitor"] 00010 00011 00012 def convert(size, box): 00013 dw = 1./size[0] 00014 dh = 1./size[1] 00015 x = (box[0] + box[1])/2.0 00016 y = (box[2] + box[3])/2.0 00017 w = box[1] - box[0] 00018 h = box[3] - box[2] 00019 x = x*dw 00020 w = w*dw 00021 y = y*dh 00022 h = h*dh 00023 return (x,y,w,h) 00024 00025 def convert_annotation(year, image_id): 00026 in_file = open('VOCdevkit/VOC%s/Annotations/%s.xml'%(year, image_id)) 00027 out_file = open('VOCdevkit/VOC%s/labels/%s.txt'%(year, image_id), 'w') 00028 tree=ET.parse(in_file) 00029 root = tree.getroot() 00030 size = root.find('size') 00031 w = int(size.find('width').text) 00032 h = int(size.find('height').text) 00033 00034 for obj in root.iter('object'): 00035 difficult = obj.find('difficult').text 00036 cls = obj.find('name').text 00037 if cls not in classes or int(difficult) == 1: 00038 continue 00039 cls_id = classes.index(cls) 00040 xmlbox = obj.find('bndbox') 00041 b = (float(xmlbox.find('xmin').text), float(xmlbox.find('xmax').text), float(xmlbox.find('ymin').text), float(xmlbox.find('ymax').text)) 00042 bb = convert((w,h), b) 00043 out_file.write(str(cls_id) + " " + " ".join([str(a) for a in bb]) + '\n') 00044 00045 wd = getcwd() 00046 00047 for year, image_set in sets: 00048 if not os.path.exists('VOCdevkit/VOC%s/labels/'%(year)): 00049 os.makedirs('VOCdevkit/VOC%s/labels/'%(year)) 00050 image_ids = open('VOCdevkit/VOC%s/ImageSets/Main/%s.txt'%(year, image_set)).read().strip().split() 00051 list_file = open('%s_%s.txt'%(year, image_set), 'w') 00052 for image_id in image_ids: 00053 list_file.write('%s/VOCdevkit/VOC%s/JPEGImages/%s.jpg\n'%(wd, year, image_id)) 00054 convert_annotation(year, image_id) 00055 list_file.close() 00056