Nazila-H / Caltech-images-convertor

This repository converts videos of caltech pedestrian dataset to images specifically for Pedestron project.

Caltech Pedestrian Dataset images-Converter

This repository is based on caltech-pedestrian-dataset-converter repository. It is impemented to prepare the images of caltech pedestrian dataset specifically for Pedestron project.

Download Caltech Datasets

1- Download caltech pedestrian videos from caltech-pedestrians.

2- The annotation files (train.json and test.json) is provided by Pedestron repository.

Virtualenv

git clone https://github.com/Nazila-H/Caltech-images-convertor.git
cd Caltech-images-convertor
conda create -name caltech python==3.8
conda activate caltech
pip install -r requirements.txt

How to ...

$ python convert_seqs.py converts .seq video files to .jpg frames that they are ready to use for Pedestron project. They will end up in the images folder.

Folder structure

convert_seqs.py
README.md
requirements.txt
data/
|__images/   # 128419 train-images
   |__set00_V000_I00000.jpg 
   |__set00_V000_I00001.jpg
   |__ ...
   |__set05_V012_I01706.jpg
or   
|__images/   # 121465 test_images
   |__set06_V000_I00000.jpg
   |__set06_V000_I00001.jpg
   |__ ...
   |__set10_V011_I01733.jpg

data/   
|__set00/  # for training 
|__set01/
|__set02/
|__set03/
|__set04/
|__set05/ 
or 
|__set06/  # for test 
|__set07/
|__set08/
|__set09/
|__set10/