Skip to content

Annotation Format Conversion

ImageNet file xml format to Darknet text format.

Full Repo here: https://github.com/campusrover/Robotics_Computer_Vision/tree/master/utils/xml_2_txt

Installation

sudo pip install -r requirements.txt

Usage

python xmltotxt.py -xml xml -out out

Example

Input xml file.

<annotation>
    <filename>image-0000016.jpg</filename>
    <size>
        <width>1920</width>
        <height>1080</height>
    </size>
    <object>
        <name>sedan</name>
        <bndbox>
            <xmin>75</xmin>
            <ymin>190</ymin>
            <xmax>125</xmax>
            <ymax>210</ymax>
        </bndbox>
    </object>
</annotation>

Output text file.

4 0.052083 0.185185 0.026042 0.018519

Motivation

I used Darknet for real-time object detection and classification. Sometimes you need to collect your own trainig dataset for train your model. I collected training dataset images and fine awesome tool for labeling images. But it generates xml files. So I needed to implement tool which translates from ImageNet xml format to Darknet text format.