Coco json converter.

This is the set of tiny python scripts to convert coco json to other format. The following formats are supported:

• Convert to XML used by Pascal VOC
• Convert to CSV used by keras-retinanet

Requirement

Python 3.6, cocoapi (pycocotools), pillow (coco_to_xml)

Usage

project
│   coco_to_csv.py
│   coco_to_xml.py
│   train.json 
│   val.json 
│
└───train
│   │   image001.jpg
│   │   image002.jpg
│   │  ...
│   
└───val
    │   image001.jpg
    │   image002.jpg
    │   ...

The directory structure described below is the same as above (train.json and the image files in train folder).

COCO to XML

coco_to_xml.py script can convert coco json to xml files. AND IT IS FAST (a few seconds for thousands of images)!

python coco_to_xml.py --coco_json=train.json --coco_folder=train --save_xml=train_xml --database_name=Database

By default, this script skip background category (class) because usually it is not required. If you want to add background category, use this flag.

python coco_to_xml.py .... --no_skip_background

Also you can check the help.

python coco_to_xml.py --help

XML Result format

The annotations xml file format is:

<annotation>
  <folder>{image_folder_name}</folder>
  <filename>{image_name}</filename>
  <source>
    <database>{database_name}</database>
  </source>
  <size>
    <width>{width}</width>
    <height>{height}</height>
    <depth>{depth}</depth>
  </size>
  <segmented>0</segmented>
  <object>
    <name>{category_name}</name>
    <pose>Unspecified</pose>
    <truncated>0</truncated>
    <difficult>0</difficult>
    <bndbox>
      <xmin>{xmin}</xmin>
      <ymin>{ymin}</ymin>
      <xmax>{xmax}</xmax>
      <ymax>{ymax}</ymax>
    </bndbox>
  </object>
</annotation>

And the xml file was saved like this:

project
│   coco_to_xml.py
│   train.json 
│   val.json 
│
└───train
│   │   image001.jpg
│   │   image002.jpg
│   │  ...
│   
└───train_xml
    │   image001.xml
    │   image002.xml
    │   ...

COCO to CSV

coco_to_xml.py script can convert coco json to single csv file.

python coco_to_csv.py --coco_json=train.json --coco_folder=train --save_ann=train.csv --save_cat=class.csv

By default, this script skip background category (class) because usually it is not required. If you want to add background category, use this flag.

python coco_to_csv.py .... --no_skip_background

And this script will check image file is exist, you can ignore that using this flag.

python coco_to_csv.py .... --no_check_file

Also you can check the help.

python coco_to_csv.py --help

CSV Result format

The annotations csv file format is:

path/to/image.jpg,x1,y1,x2,y2,class_name

The classes csv file format is:

class_name,id

License

MIT License