The QCNode(Qualcomm Compute Node) is Qualcomm ADAS hardware abstraction layer that provides user friendly APIs to access the Qualcomm ADAS hardware accelerator. It is ADAS service oriented to provide user friendly node and utils to cover the perception and compute vision related task.
Refer to the INDEX for full documentation.
Below is a summary information of the current implemented nodes.
| QCNode | QNX PVM | HGY PVM | Library | Processor | Features |
|---|---|---|---|---|---|
| Camera | YES | YES | qcarcam | Camera | camera streaming |
| Remap | YES | YES | FastADAS | CPU, HTP0, HTP1, GPU | Color Conversion, ROI Crop, Downscaling, Undistortion |
| QNN | YES | YES | QNN | CPU, GPU, HTP0, HTP1 | AI model inference |
| VideoEncoder | YES | YES | vidc | VPU | Encode image to video frame |
| VideoDecoder | YES | YES | vidc | VPU | Decode video frame to image |
| Voxelization | YES | YES | FastADAS, OpenCL | CPU, HTP0, HTP1, GPU | Generate pillars from point clouds |
| CL2DFlex | YES | Yes | OpenCL | GPU | Color Conversion, Resize, ROI Crop, Undistortion |
| OpticalFlow | YES | YES | SV | EVA | Optical flow computation |
| DepthFromStereo | YES | YES | SV | EVA | Depth estimation from stereo images |
For how to build the QCNode package manually, check this:
- HGY Linux: How to build for HGY Linux
- HGY Ubuntu: How to build for HGY Ubuntu
For how to run the QCNode package, check this README.
For how to run the QCNode E2E perception pipeline sample, check this README.
QCNode is licensed under the BSD-3-clause License. See LICENSE.txt for the full license text.