ImagineS (Image Engine Search) tool augments an image dataset by searching a series of queries in Google Images.
-
We recommend first creating an environment by using either Virtual Envs or Conda.
-
Then install the library using
pip:pip install imagines
Example: Download apple and water related images by adding the query searches that would belong to each class.
from imagines import DatasetAugmentation
dataset_augmentation = DatasetAugmentation(
driver_type='chrome'
)
label_queries = {
'apple': ['apple'],
'water': ['water', 'water bottle'],
}
max_links_to_fetch = 3
sleep_between_interactions = 1
image_shape = (224, 224)
resize_images = True
output_dir = 'images/'
dataset_augmentation.augment_dataset(
label_queries,
output_dir,
max_links_to_fetch,
image_shape,
resize_images,
sleep_between_interactions
)