search

irolaina / FCRN-DepthPrediction

Deeper Depth Prediction with Fully Convolutional Residual Networks (FCRN)
BSD 2-Clause "Simplified" License
1.12k stars 312 forks source link

NYU Depth v2 Preprocessing #64

Closed nicolasrosa closed 6 years ago

nicolasrosa commented 6 years ago

Did you train the model on the Kinect's raw depth images (uint16) OR Did you implement the following calculations in your preprocessing routine? Did you also swap the bytes from the depth image?

imgDepthOrig: swapbytes(kinect_raw_depth_image_uint16) imgDepthAbs: Absolute depth in meters*

*I believe this will be between 0~10m. Is that correct?

Code: depthParam1 = 351.3; depthParam2 = 1092.5; imgDepthAbs = depthParam1 ./ (depthParam2 - imgDepthOrig);

Matlab Script from dataset's toolkit: depth_rel2depth_abs.txt

Ariel-JUAN commented 6 years ago

Yes, it is between 0~10m.