xview_babyversion

  1. make sure you have the same folder struct

  2. download the geojson files, and baseline folder from xview organization. download the train, val, train_label data. Below is the folder structure I used.

      ./xview

            './train_images'
            './val_images'

            './train_labels/xView_train.geojson'

            './baseline/xview_class_labels.txt'

            './scripts/*'

  1. cd folder 'script'

  2. run the command below. It will parse the label json file, and generate an annotation file and a classes file in csv format. They are saved in the same local direcotry.

python generate_labels.py

  The generated annotation file format looks like : (if a image does not contain any target object, and you wanna train the as the negatives, their bouding box will be empty, as the last line shows.

/data/imgs/img_001.jpg,837,346,981,456,cow
/data/imgs/img_002.jpg,215,312,279,391,cat
/data/imgs/img_002.jpg,22,5,89,84,bird
/data/imgs/img_003.jpg,,,,,