Martin Kersner, m.kersner@gmail.com
Caltech Pedestrian Detection Benchmark is stored in two exotic formats. Images are stored in seq format and labels are in vbb. In order to be able to work with such formats one has to be able to access original data. There is support for Matlab with Piotr's Computer Vision Matlab Toolbox, however there is no full support for both formats in Python.
This repository consist of two scripts, one (extract_annotations.py) for extracting information from seq format and one (extract_images.py) for vbb format.
extract_annotations.py was taken from mitmul's repository and slightly modified.
extract_images.py was inspired by jainanshul's repository and Piotr's Computer Vision Matlab Toolbox.
git clone https://github.com/martinkersner/caltech-pedestrian-detection-benchmark-python-extractor.git
Creates XML file called annotations.xml
python extract_annotations.py "path/to/directory/with/set/directories"
python extract_images.py "path/to/directory/with/seq/files" "path/for/output/images"