This is a tool to convert various datasets into .raw format that fits kfusion and ofusion.
To convert a ViSim dataset into .raw format, please put images into the following structure:
Given a folder dataset:
dataset/cam0/data -- where you put the rgb images
dataset/cam0/data.csv -- where you put a csv file that contains filenames of all rgb images
dataset/depth0/data -- where you put the depth images
dataset/depth0/data.csv -- where you put a csv file that contains filenames of all depth images
dataset/cameraInfo.txt -- where you put intrinsic of the camera
A sample cameraInfo.txt could look like this:
#Image Res.(w,h):
640 480
#focus lens :
606.0509716312624 609.2072669938981
#principal pt :
327.88802413335253 239.04273996525234
#Undist. coeff. :
0 0 0
Then please run:
./vs2Raw [path-to-ViSim-dataset] [camera-index] [optional: low-quality-image (0 or 1)]
To convert a Co-fusion synthetic dataset into .raw format, please put images into the following structure:
Given a folder dataset:
dataset/colour -- where you put the rgb images
dataset/depth_original -- where you put the depth images
dataset/trajectories/gt-cam-0.txt -- where you put the GT camera trajectory
dataset/calibration.txt -- where you put intrinsic of the camera
A sample calibration.txt could look like this:
564.3 564.3 480 270 960 540
which follows format:
[focal lens x] [focal lens y] [image center x] [image center y] [image width] [image height]
Then please run:
./cf2Raw [path-to-ViSim-Folder] [optional: use-noisy-depth (0 or 1)]
Then please run:
./klg2Raw [path-to-klg-file] [image width] [image height] [image center x] [image center y] [focal lens x] [focal lens y]