If you are familiar with the data set you can immediately download the entire data set via links at the bottom of this page. Otherwise you may find the evaluation section below to be a useful introduction.
(*)These data have been subsampled (by varying amounts, depending on the length of the sample); they require Flash to view.
(**)Created using the front facing camera on the Ladybug Camera.
(***)Images taken directly from the camera – to undistort and rectify them (before stereo matching for example), please see the code provided.
(****)To view the X3D point cloud requires a X3D viewer; FreeWrl is available for Windows, Linux, macOS and Android.
Extracting Data From The Plain-text “alog files”
All non-image data is stored in a plain-text log file called an “alog”. To get acquainted with this format and suitable access methods we suggest you:
- Download some data evaluation snippets above, particularly the alog which is the text file containing all non-image data, and the Flash Laser Viewer. The data set is large and users may not wish to download 30G data and start using the parsing tools immediately so we have provided some short partitions of the alog file.
- Download and compile (as a standalone executable or from within Matlab©) the alog parsing tool we provide called uAlogParser. Full instructions on how to compile this single source file on multiple platforms are given here.
- Try running the parser on one of the smaller evaluation alog segments.
- If you have access to Matlab©, try running the point cloud builder on an evaluation chunk to see some 3D laser data.
- Download the full alog file – perhaps after finding an event that appears interesting.
Full Data Set
Here we provide the data set in full. The data set consists of three streams of information: two sets of images, one from the Stereo Camera, and one from the Ladybug Camera, and an Alog file which contains other information like laser data and odometry. The image sets come in two formats. Firstly, we provide them as single gzip’ed tarballs (.tgz) containing all the images, via ftp. These are large files due to the number of images, so the second format is the image sets broken down into ~500MB chunks (via http).
Image Sets, Single Tarball
(ftp) Stereo Images (16G) (MD5SUM) (*) (**)
(ftp) Ladybug Images (9.1G) (MD5SUM) (*)
(ftp) Ladybug Panoramas (1.3G) (MD5SUM) (*)
Image Sets, Multiple Tarballs/Zip Archives
We suggest a download manager such as DownTHEMall! (Mozilla only) to download the 500MB chunks painlessly.
(http) Stereo Images (**) (.tgz)
(http) Ladybug Images (.tgz)
(http) Panorama Images(***) (.zip)
(http) Alog File (220M) (*)
Visual Odometry Data[14-06-10] Please note that this data was not originally part of the IJRR submission.
(http) Visual Odometry (VO) data (7M)
(*)Please note that files greater than 4GB are located on anonymous public ftp. If you are prompted for a username and password, use the following:
(**) Please note that the stereo images contain lens distortion and are unrectified. This makes them unsuitable for direct use in, for example, local window-based stereo matching. Code is provided that will perform undisortion and rectification on these image sequences. Download details and instructions for use are given: Code to undistort and rectify stereo image sequence.
(***) Panoramas are provided as a quick overview of the ladybug image data. Each image is a concatenation of the 5 camera images, all captured at the same time. We suggest using the raw ladybug images as these correspond to the files referred to in the .alog file.