Extract images and annotation files from the Caltech Pedestrian Dataset using python.

This script extracts and converts data from .seq and .vbb files to .jpg images and .json files. Individual images + annotation files are extracted and stored per sequence and per video for easier access.

• python>=2.7 or python>=3.3
• numpy
• scipy
• json

In order to start using this script you must first untar all files before running the script.

Then, to extract files just run the following comand in a terminal:

python converter.py -data_path <path_to_data>

This will create a dir named extracted_data/ and store all extracted data in there. If you wish to specify a different directory to store your data you can set the -save_path input argument to a different path.

python converter.py -data_path <path_to_data> -save_path <path_to_extract>

Note: A valid -data_path is required and must be provided in order for the script to work.

Licensed under the MIT license.

This code is based on @hizhangp caltech converter code.