Kitty-stereo-dataset-parser

This parser is build upon pykitty repo made by utiasSTARS available at: utiasSTARS/pykitti. I decided to build my own parser for evaluation of tradition stereo vision technique which is available in opencv library. Just for simplicity reasons I incorporated whole pykitty repo in my own repo. This way anyone who would like to use this repo won't need to download pykitty.

Features

• Right and left rectified images from both views of stereo vision system
• Ground truth map generated from lidar data transformed to image plane
• Distance map generated based on disparity map produced by stereo vision system

Data

First step is to download data from kitty site. You can choose whichever dataset you want (it should not matter). For example purpose we will use dataset labeled as "2011_09_26_drive_0001". Download synced+rectified data and calibration data and unzip it. Move both uziped folders to repo's data folder.

Repo's stucture tree should look like:

Stereo
│   README.md
│   LICENCE
│   ...
│───data
│       │─── 2011_09_26
│           │   calib_cam_to_cam.txt
│           │   calib_imu_to_velo.txt
│           │   calib_velo_to_cam.txt
│           │─── 2011_09_26_drive_0001_sync
│───pykitty
│       │   ...

Install dependencies

Create and activate virtual environment:

conda create -n yourenvname python=3.7

conda activate yourenvname

Go to root and install needed dependencies:

pip install -r requirements.txt

How to use

Go to source/parse_data.py and run. The script should produce an image shown bellow:

Acknowledgment

Thank you for good work! https://github.com/utiasSTARS/pykitti