UA-DETRAC to COCO

The following code converts the UA-DETRAC data to the COCO format. You will need

1. xml_to_coco.py
2. pycoco_visualize.py
3. cocotools.yml

Installing the dependencies

I assume you are using the conda package environment. cd into the directory that contains the cocotools.yml. Then run

conda env -f cocotools.yml

These are all the dependencies you need

Running the code

The organization of the folder must be very specific, as follows:

project
│   README.md
│   xml_to_coco.py
│   pycoco_visualize.py
│   MVI_20032.xml
└───MVI_20032
│   │   file011.jpg
│   │   file012.jpg
│   
│   MVI_20035.xml
└───MVI_20035
    │   file01.jpg
    │   file02.jpg

Basically, the folders with all the images, the XML files and xml_to_coco.py and pycoco_visualize should be in the same folder

Running the code

Activate the environment using conda activate cocotools. The code is run with the following command

python xml_to_coco.py --visualize 1

When the --visualize flag is set to one, you will see 5 random samples of the annotations overlayed on the original images. The annotations are written to annotations.json in the same folder as above.