Alexey
2018-05-17 e39d1011095de8e7428fe0e76b73986db3ca6918
Merge pull request #844 from salbatron/patch-1

Create get_openimages_dataset.py
1 files added
19 ■■■■■ changed files
scripts/get_openimages_dataset.py 19 ●●●●● patch | view | raw | blame | history
scripts/get_openimages_dataset.py
New file
@@ -0,0 +1,19 @@
import csv
import os
#select classes you want to download at https://github.com/openimages/dataset/blob/master/dict.csv
CLASS_LIST = ('/m/01g317','/m/04yx4')
img_name = "111111111111"
#download csv from https://storage.googleapis.com/openimages/web/download.html
with open('path\\train-annotations-bbox.csv', newline='') as csvfile:
    bboxs = csv.reader(csvfile, delimiter=',', quotechar='|')
    for bbox in bboxs:
        if bbox[2] in CLASS_LIST:
            if img_name != bbox[0]:
                if not os.path.isfile("destination_path\\%s.jpg"%bbox[0]):
                    os.system("gsutil cp gs://open-images-dataset/train/%s.jpg destination_path"%bbox[0])
                    out_file = open("destination_path\\%s.txt"%bbox[0], 'w')
                    img_name = bbox[0]
            if img_name == bbox[0]:
                out_file.write(str(CLASS_LIST.index(bbox[2])) + " " + str(float(bbox[4])+(float(bbox[5])-float(bbox[4]))/2) + " " + str(float(bbox[6])+(float(bbox[7])-float(bbox[6]))/2)+ " " + str(float(bbox[5])-float(bbox[4])) + " " + str(float(bbox[7])-float(bbox[6])) + '\n')