Stereo depth images for the car instance dataset

[jingnanshi]

Hi,

The images in the stereo dataset overlap almost 100% with those in the car instance dataset. Unfortunately, due to cropping, such stereo images cannot be directly used in the car instance recognition experiment.

I'm wondering whether the parameters used for rectification/undistortion/cropping can be provided for the stereo dataset, so that I can apply the same procedure to the images in the car instance dataset and use the depth obtained for car instance recognition experiments.

Thanks in advance!