samples/imageprocessing/stereo/stereoPVA/README.md

Go to the documentation of this file.

# Copyright (c) 2018-2019, NVIDIA CORPORATION.  All rights reserved.

@page dwx_stereo_disparity_pva_sample Stereo Disparity PVA Sample
@tableofcontents

@section dwx_stereo_disparity_pva_description Description

The Stereo Disparity sample demonstrates the stereo pipeline running on multiple hardware engines, namely:

- CPU: Input images are converted from YUV420 Semiplanar to Y for the stereo pipeline.
- VIC: Input images are transformed according to the requirements of the engines.
- PVA: Hints are generated in the preprocessing step, and the output is refined in the postprocessing step.
- NVENC: Stereo disparity is computed.
- GPU: Operations are performed on the output to get the disparity values.

The sample reads frames from two stereo videos and rectifies them on the basis
of the camera calibration contained in the rig configuration file. Then, it runs the
frames through the stereo pipeline mentioned above and displays final stereo
output.

@section dwx_stereo_disparity_pva_running Running the Sample

The command line for the sample is:

    ./sample_stereo_disparity_pva --rigconfig=[path/to/rig/configuration/file]
                                  --video-left=[path/to/the/left/video]
                                  --video-right=[path/to/the/right/video]
                                  --sensor-left=[name/of/left/camera/in/rig/file]
                                  --sensor-right=[name/of/right/camera/in/rig/file]
                                  --maxDistance=[maximum/distance/in/meters/for/depth/rendering]
                                  --estimationMode=[LQ|MQ|HQ]
                                  --nvencNo=[0|1]
                                  --pvaNo=[0|1]

where

    --rigconfig=[path/to/rig/configuration/file]
        Points to the rig file that contains the intrinsics and extrinsics for the stereo camera.
        Default value: path/to/data/samples/stereo/full.json

    --video-left=[path/to/the/left/video]
        Is the path to the video recorded from the left camera.
        Default value: path/to/data/samples/stereo/left_1.h264

    --sensor-left=[name/of/left/camera/in/rig/file]
        Is the name of the left input sensor in the rig file.
        Defaul value: left_60FOV

    --sensor-right=[name/of/right/camera/in/rig/file]
        Is the name of the right input sensor in the rig file.
        Defaul value: right_60FOV

    --video-right=[path/to/the/right/video]
        Is the path to the video recorded from the right camera.
        Default value: path/to/data/samples/stereo/right_1.h264

    --maxDistance=[maximum/distance/in/meters/for/depth/rendering]
        Maximum distance in meters for depth rendering.
        Every pixel with a depth greater than maxDistance is rendered as blue
        Default value: 40.0

    --estimationMode=[LQ|MQ|HQ]
            Indicates the quality and the performance of Stereo algorithm:
            - LQ: Low Quality
            - MQ: Medium Quality
            - HQ: High Quality
            Default value: MQ

    --nvencNo=[0|1]
            Indicates on which Nvidia Video Encoder the corresponding stage of the pipeline is to be executed.
            Default value: 0

    --pvaNo=[0|1]
            Indicates on which PVA engine the corresponding stage of the pipeline is to be executed.
            Default value: 0

@section dwx_stereo_disparity_pva_output Output

The sample creates a window and displays the disparity view composed by:

- Left: Anaglyph of left and right image
- Right: Stereo disparity

The stereo output is color coded for clarity.

![Stereo Disparity on NVENC & PVA](sample_stereo_disparity_pva.png)

@section dwx_stereo_disparity_pva_more Additional information

For more details see @ref stereo_mainsection .