lih627/dataset-utils

Dataset-Utils

This repository is created to process data labels in the field of detection,

voc2coco contains function which can convert pascal voc .xml to coco .json format.

datasetutils can transform images with annotations to low resolution images with annotations.

voc2coco

You can get the image annotations by labelImg.

It's usually Pascal VOC format. The file directory is as follows:

MarkerDataset
├── annotations
│   ├── 000000.xml
│   ├── 000001.xml
│   ├── ...
│   └── ...
├── images
│   ├── 000000.jpg
│   ├── 000001.jpg
│   ├── ...
│   └── ...
├── coco_output.json # convert function output this file
├── except.txt # contains some xml which don't have boox
└── xmllist.txt # contains all xml file names in annotations dir

To convert voc to coco, we can use:

python voc2coco.py --xml_path /path/to/annotaions --xml_list_path /path/to/xmllist.txt --output_path /path/to/coco_output.json

It will create xmllist.txt and except.txt which contain the name of xml files, and coco_output.json is coco format annotations.

All these are used for my detection dataset

datasetutils

This file can transform high resolution images with annotations to low resolution images.

For example, My dataset have 25,000 images with annotations. The resolution is 1280×720. Use this python file to create a low resolution dataset. It contains 25,000 images with annotations, but resolution is changed to 640×360. It also changed the xml annotation file to match low resolution images.

dataset-utils's Issues

Filename Error

HI @lih627 I wanted to ask that how can I work around your code if I want to use a string as a file name and not an integer for example "some_file_name_0.xml" as well as "some_file_name_0.png"?