optical.converter.utils

__author__: HashTagML license: MIT Created: Sunday, 28th March 2021

Functions

copyfile(src, dest[, filename])	copies a file from one path to another
create_tf_example(df, root)	returns protobuf for a given image
exists(path)	checks for whether a directory or file exists in the specified path
filter_split_category(df[, split, category])	given the label df, filters the dataframe by split and/or label category
find_job_metadata_key(json_data)	finds metadata key for sagemaker manifest format
find_splits(image_dir, annotation_dir, format)	find the splits in the dataset, will ignore splits for which no annotation is found
get_annotation_dir(root)	returns annotation directory given a root directory
get_id_to_class_map(df)	This function return the class_id to class name mapping
get_image_dir(root)	returns image directory given a root directory
ifnone(x, y[, transform, type_safe])	if x is None return y otherwise x after applying transofrmation transform and casting the result back to original type if type_safe
read_coco(coco_json)	read a coco json and returns the images, annotations and categories dict separately
read_xml(xml_folder, img_path)	read xml files in the folder and return list's of information used to construct master_df
tf_decode_image(root, data, split)	Decodes images and save in images folder under root
write_json(data_dict, filename)	writes json to disk
write_label_map(id_to_class_map, output_dir)	writes label_map used in tf object detection
write_xml(df, image_root[, output_dir])	write xml files in PASCAL VOC format given a label dataframe