Closed aasharma90 closed 6 years ago
lyndonzheng
commented
6 years ago Hi @aasharma90 For VKITTI, we just split the trainA_SYN.txt and testA_SYN.txt for depth estimation. For image-to-image translation video, we choice the very similar scene in VKITTI and KITTI. In this txt file, 03030079_A.png, (03-03-0079) represents the '0001/15-deg-left/00079.png', and 07120772_A.png (07-12-0772) represents the '0020/sunset/00772.png'. For the KITTI, just use the standard split.
aasharma90
commented
6 years ago Hi @lyndonzheng Does it affect your results if you do not have similar scenes from the two datasets for your translation network? I am currently experimenting only with the translation network, and was wondering if I can just use all the images in VKITTI (not in any particular order) v/s all the images in *datasplit/kitti_train_files.txt file?
Actually, I intend to use your code for some other datasets, and was wondering whether similarity is really a requirement? For instance, is it possible to map SYNTHIA into KITTI?
Saurav-31
commented
2 years ago Hi @aasharma90 For VKITTI, we just split the trainA_SYN.txt and testA_SYN.txt for depth estimation. For image-to-image translation video, we choice the very similar scene in VKITTI and KITTI. In this txt file, 03030079_A.png, (03-03-0079) represents the '0001/15-deg-left/00079.png', and 07120772_A.png (07-12-0772) represents the '0020/sunset/00772.png'. For the KITTI, just use the standard split.
Hi @lyndonzheng, How to interpret the codes for vkitti dataset
<world>_<variation>_<image_no> ??
What I mean is how to know 03 means 15-deg-left? Can you please point me to the that. Thanks in advance.
Hi @lyndonzheng Is it possible for you to provide us a list of your (train, val, test) datalists for both VKITTI and KITTI datasets respectively? Maybe for the latter, they are *datasplit/kitti_*.txt files? Could you please confirm?