First of all, thank you for making this wonderful project available.
I'm planning to use this algorithm in my thesis, for which I need to get real depth, which will be consistent with camera poses from the same dataset ( I use Kitti-stereo-2015). I've tried to estimate depth from LEAStereo disparity using the formula:
depth = fx * baseline / disparity
However when I combine the resulting depth with camera extrinsics to reproject pixels from one frame to another (for example from left frame at time t to left frame at time t+1), the results aren't meaningful (some pixel values turn out to be -40000 which is way beyond image size).
I'd like to know what is the correct formula to get real depth from the outputted disparity, I'll appreciate any help or insights. Thank you!
First of all, thank you for making this wonderful project available.
I'm planning to use this algorithm in my thesis, for which I need to get real depth, which will be consistent with camera poses from the same dataset ( I use Kitti-stereo-2015). I've tried to estimate depth from LEAStereo disparity using the formula:
depth = fx * baseline / disparity
However when I combine the resulting depth with camera extrinsics to reproject pixels from one frame to another (for example from left frame at time t to left frame at time t+1), the results aren't meaningful (some pixel values turn out to be -40000 which is way beyond image size).
I'd like to know what is the correct formula to get real depth from the outputted disparity, I'll appreciate any help or insights. Thank you!