Export dataset to VSRS

This is how to export an entire dataset (images and depths) to VSRS. It must first have been imported (for example, from Kinect), and contain the reprojected depth maps.

1. Prepare dataset parameters

In the dataset parameters, the image_filename_template and depth_filename_template in the root group must point to existing PNG images and reprojected depth maps. Add a vsrs group, and put in the file names of the YUV files for images and disparity maps, that should be generated.

Also inside the vsrs group, add two numbers z_near and z_far. They should indicate the minimal and maximal depth values in the depth maps, and are used for the convertion from depth map to disparity map. (See vsrs/vsrs_disparity). Good values can be found by using misc/view_depth on one of the depth maps.

2. Run the export script

Then, run

vsrs/export_for_vsrs.py parameters.json

If the environment variable LICORNEA_PARALLEL is set to 1, it will run in parallel, like with the import from Kinect script.

3. Export camera parameters

To convert the camera parameters file to VSRS format, use

camera/export_mpeg cameras.json cameras_vsrs.txt