Monday, November 19, 2012

1

THE NEW TSUKUBA STEREO DATASET IS OUT!

Finally! After so much work we released the "New Tsukuba Stereo Dataset" at ICPR2012


This dataset contains 1800 stereo pairs with ground truth disparity maps, occlusion maps and discontinuity maps that will help to further develop the state of the art of stereo matching algorithms and evaluate its performance. It has been generated using photo-realistic computer graphics techniques and modeled after the original "head and lamp" stereo scene released by University of Tsukuba in 1997. 

The dataset is a 1 minute video sequence and also contains the 3D position and orientation of the camera on each frame, so it also can be used to develop and evaluate camera tracking methods.

It can be downloaded freely from the CVLAB website

Enjoy it!

1 comment:

Diana said...

Hello, Martin!

I`m working on a project that involves calibrating a stereo system. I read the posts on your blog and they really helped me a lot in understanding the involved concepts and writing the code. But still I get very bad calibration (and rectification) results and I would like to ask you for a piece of advice.

My system is a vertical stereo system and one camera is slightly facing the other camera (there is an angle of about 10 degrees between the cameras). Also, the system has a (fixed) zoom, so the calibration pattern that I used has a small square size (each square is 1cm width and 1cm height).

The calibration and the rectification code that I wrote works good if I use calibration images (images of the chessboard) taken from the internet (with your samples and those provided by OpenCV), but gives bad results using the images taken with my stereo system (the principal point is no were near the center of the image, the baseline is wrong etc.). I took about 10 sets of images (14 images on each set) with my system and the results are still bad.

I would like to ask you if this is because of the parameters of the stereo system (cameras not parallel, the zoom, the calibration pattern is too small) or I`m doing something wrong when capturing the calibration images.

Thank you in advance!
Best regards,
Diana

Post a Comment

Facebook

Twitter

Friends