VIA2COCO

This repository contains Python code to convert annotations from Oxford's VGG Image Annotator (VIA) to Microsoft's Common Objects in Context (COCO) format.

Installation

• Install the latest version of Python 3.X.

Usage

The simplest usage would be to specify a list of categories.

Caveat:

• categories should have distinct names,
• the order matters, as it will be used to number categories.

import convert as via2coco

input_dir = '/content/data/balloon/train/'
input_json = input_dir + 'via_region_data.json'
categories = ['balloon']

coco_dict = via2coco.convert(
    imgdir=input_dir,
    annpath=input_json,
    categories=categories,
)

By default, the first category is indexed as 1. This can be changed to 0 (or any non-negative integer) as follows:

import convert as via2coco

input_dir = '/content/data/balloon/train/'
input_json = input_dir + 'via_region_data.json'
categories = ['balloon']
first_class_index = 0

coco_dict = via2coco.convert(
    imgdir=input_dir,
    annpath=input_json,
    categories=categories,
    first_class_index=first_class_index,
)

To save the output as a JSON file:

import convert as via2coco

input_dir = '/content/data/balloon/train/'
input_json = input_dir + 'via_region_data.json'
categories = ['balloon']

output_json = input_dir + 'coco_train.json'

coco_dict = via2coco.convert(
    imgdir=input_dir,
    annpath=input_json,
    categories=categories,
    output_file_name=output_json,
)

It is also possible to specify a list of super-categories.

Caveat:

• super-categories can have the same name,
• the lists of categories and super-categories should have the same length,
• the order matters, as it will be used to match categories and super-categories.

import convert as via2coco

input_dir = '/content/data/balloon/train/'
input_json = input_dir + 'via_region_data.json'
categories = ['balloon']

super_categories = ['N/A']

coco_dict = via2coco.convert(
    imgdir=input_dir,
    annpath=input_json,
    categories=categories,
    super_categories=super_categories,
)

References

• VIA2COCO: the original repository
• VGG Image Annotator (VIA), by Oxford's Visual Geometry Group
• Common Objects in Context (COCO), by Microsoft
• An example of the VIA format, as shown in a Github issue