I did the alignment between color and depth image(Azure kinect dk) with CUDA. It takes so long time because it is necessary to tranform the 2D-3D conversion. I tested this on Windows 10. If you wanna use this on Linux, you could compile this to nvcc without cmake (with dependency).
The code provided by Azure kinect SDK can only use BGRA32 which is one of kind of image format. This code can use other image formats like MJPG.
- Azure kinect : <Color : [1280 x 720], Depth : [640 x 576]>
- OS : Windows 10 (NOT Linux)
- IDE : Visual studio 2015 community
- CPU : Intel(R) Core(TM) i7-9700K (3.60GHz)
- GPU : Geforce RTX 2080 ti
- RAM : 64 GB
- Opencv : 4.3.0 (just for visualizing)
- Azure kinect SDK
- CUDA : 10.1
