Image Annotation Tools

Features

1. Double Check IoU Annotation Tool

in ./double_check_iou_tool folder

Double check 2 coco json. Just for bbox, not for segmentation.

Keep the bbox with IoU(Intersection over Union) greater than the threshold among the different two group of annotations from two people in the same image.

The remaining annotations require further manual review.

# how to run it

e.g.
    
coco_json1_file_path = 'J:\\workspace\\xxxxxx\\dataset1-annotator1.json'
coco_json2_file_path = 'J:\\workspace\\xxxxxx\\dataset1-annotator2.json'

double_check_two_cocos(coco_json1_file_path, coco_json2_file_path, os.path.join(os.path.abspath(os.curdir), "results"))

<!--Groups represent the two annotators.-->
<!--Results contains two group folders and one conclusion folder.-->
<!--Each group have matched and suspected annotations.-->

<!--Double-check result will save in conclusion folder.-->
<!--when one annotaion is matched with another annotation, the matched annotation with smaller area box will be saved in matched conclusion folder.-->
<!--The final conclusion merge all matched annotations and suspected annotations with different color.-->
<!--# color green represent matched. #1A971E-->
<!--# color red represent not matched. #C9151A-->

<!--Results is a batch of coco json files of one labeled image. For convenient import, conclusion result's metadata is based on group1 image, including: image id/file_name/file_path, and based on group1's category id.-->

# result file structure:

results/
    conclusion/
        all/
            image1.json  # merged all matched and suspected annotations
            image2.json
        matched/
            image1.json
            image2.json
        suspected/
            image1.json
            image2.json
    group1/
        matched/
            image1.json
            image2.json
        suspected/
            image1.json
            image2.json
    group2/
        matched/
            image1.json
            image2.json
        suspected/
            image1.json
            image2.json

2. COCO Json Merge / Split Tool

Split Tool: Split the coco json of one dataset with multiple labeled images, to a batch of coco json files of one labeled image.

in ./coco_json_merge_split_tool folder

# Split coco json
# how to run it
split_coco(input_file_path, output_dir_path)

Merge Tool: Merge a batch of coco json files of one labeled image to one coco json of one dataset with multiple labeled images.

# Merge coco json
# how to run it

<!--Merge the files in input dir, each file is a coco json file for one image.-->
input_dir_path = "J:\\workspace\\xxxxxx\\splited_double_check"

<!--Merge coco json result will be saved in this file-->
output_file = "J:\\workspace\\xxxxxx\\merged_double_check\\merged_double_check.json"

<!--The output file's image id and category id will be the same as the metadata file, mapping the same image filename and category name.-->
coco_metadata_json_file = "J:\\workspace\\xxxxxx\\dataset1-metadata.json"

<!--Optional Param. If you use the original dataset with same image id, you can set this file path to "".-->
<!--If you want to import the result to another dataset, you need to remap the image id, you can use this file path.-->
<!--This image id mapping will reset the image id with the same image file_name. is a image json array in coco image rest api format.-->
<!-- file content e.g.
        [
            {
            "id": 2030,
            "file_name": "Bthe_01.jpg",
            "annotated": false,
            "annotating": [],
            "num_annotations": 0
            },
            ...
        ]
-->
image_id_remapping_file = "J:\\workspace\\xxxxxx\\double_check_dataset_image_mapping.json"

merge_coco(input_dir_path, output_file, coco_metadata_json_file, image_id_remapping_file)