ply blurry

[flyisme]

win10(cuda10.2 )+GeForce GTX 1060+vs 2017+ intelD415

[flyisme]

Is there a problem with the parameter configuration? How should I deal with this? Any suggestions? Thank you

[tyronedong]

Glad to see you are running this on d415. I am using d435 but haven't run seccessfully. Could you please share your d415 driver code. It would be great if I could have you WeChat.

[flyisme]

Glad to see you are running this on d415. I am using d435 but haven't run seccessfully. Could you please share your d415 driver code. It would be great if I could have you WeChat.

Configure the 2.0sdk.

#include "stdafx.h"
#include "librealsense2/rs.hpp"
#include "RealSenseSensor2d0.h"

#ifdef REAL_SENSE

RealSenseSensor2d0::RealSenseSensor2d0()
{
    //d515
    /*unsigned int depthWidth = 1024; //! todo
    unsigned int depthHeight = 768;
    unsigned int colorWidth = 1920;
    unsigned int colorHeight = 1080;*/

    //d435i d415
    unsigned int depthWidth = 1280; //! todo
    unsigned int depthHeight = 720;
    unsigned int colorWidth = 1920;
    unsigned int colorHeight = 1080;

    RGBDSensor::init(depthWidth, depthHeight, colorWidth, colorHeight);

    //RGB
    cfg.enable_stream(RS2_STREAM_COLOR, colorWidth, colorHeight, RS2_FORMAT_RGB8, 30);
    //DEPTH
    cfg.enable_stream(RS2_STREAM_DEPTH, depthWidth, depthHeight, RS2_FORMAT_Z16, 30);
}

RealSenseSensor2d0::~RealSenseSensor2d0()
{

}

void RealSenseSensor2d0::createFirstConnected()
{
    // Start streaming with the default recommended configuration
    rs2::pipeline_profile pipe_profile =pipe.start(cfg);

    const int fisheye_sensor_idx = 1;

    rs2::stream_profile fisheye_stream = pipe_profile.get_stream(RS2_STREAM_DEPTH, 0);
    rs2_intrinsics intrinsics = fisheye_stream.as<rs2::video_stream_profile>().get_intrinsics();

    printf("depth:fx:%f,fy:%f\n", intrinsics.fx,intrinsics.fy);

    //Éî¶ÈÏà»úÄÚ²Î
    initializeDepthIntrinsics(intrinsics.fx, intrinsics.fy, getDepthWidth() / 2.f, getDepthHeight() / 2.f);

    fisheye_stream = pipe_profile.get_stream(RS2_STREAM_COLOR, 0);
    intrinsics = fisheye_stream.as<rs2::video_stream_profile>().get_intrinsics();

    //RGBÏà»úÄÚ²Î
    initializeColorIntrinsics(intrinsics.fx, intrinsics.fy, getColorWidth() / 2.f, getColorHeight() / 2.f);

    printf("color:fx:%f,fy:%f\n", intrinsics.fx, intrinsics.fy);

    //Ïà»úÍâ²Î
    initializeColorExtrinsics(mat4f::identity());
    initializeDepthExtrinsics(mat4f::identity());
}

bool RealSenseSensor2d0::processDepth()
{
    rs2::frameset data = pipe.wait_for_frames(); // Wait for next set of frames from the camera
    rs2::depth_frame depth = data.get_depth_frame();
    rs2::video_frame color = data.get_color_frame();            // Find the color data

    /*rs2::colorizer c;
    color = c.colorize(depth);*/

    rs2_error* e = 0;

    printf("\nframe num:%d,",data.get_frame_number());

    float* depth_f = getDepthFloat();
    //float* depth_f  =new float[getDepthWidth()*getDepthHeight()];

    //const uint16_t* depth_frame_data = (const uint16_t*)(rs2_get_frame_data(depth.get(), &e));

    //const uint16_t* depth_frame_data = (const uint16_t*)(depth.get_data());

    vec4uc* color_v = getColorRGBX();
    //vec4uc* color_v= new vec4uc[getColorWidth()*getColorHeight()];

    int n = getColorHeight();

    const uint8_t* rgb_frame_data = (const uint8_t*)(color.get_data());
    int stride = color.get_stride_in_bytes();

    for (int i = 0; i < (int)getColorWidth(); i++)
    {
        for (int j = 0; j < (int)getColorHeight(); j++) {

            //R
            int R= rgb_frame_data[j * stride + (i * 3)];
            //G
            int G=rgb_frame_data[j * stride + (i * 3) + 1];
            //B
            int B=rgb_frame_data[j * stride + (i * 3) + 2];
            color_v[i+j* getColorWidth()] = vec4uc(R,G, B, 255);
        }
    }

    printf("ÖÐÐĵã¾àÀ룺%f\n", depth.get_distance(getDepthWidth()/2, getDepthHeight()/2));

    for (int x = 0; x < getDepthWidth(); x++)
    {
        for (int y = 0; y < getDepthHeight(); y++)
        {
            //const USHORT& d = *depth_frame_data++;

            depth_f[x + y * getDepthWidth()]= depth.get_distance(x,y);

            /*if (d == 0)
            {
                depth_f[x + y * getDepthWidth()] = 0.0f;
            }
            else
            {
                depth_f[x + y * getDepthHeight()] = (float)d;
            }*/

        }
    }

    // depth
    //memcpy(m_depthFloat[m_currentRingBufIdx], depth_f, sizeof(float)*getDepthWidth()*getDepthHeight());

    // color
    //memcpy(m_colorRGBX, color_v, sizeof(vec4uc)*getColorWidth()*getColorHeight());
    return true;
}

#endif

[tyronedong]

Thanks for your sharing! Now I have a new problem. I compile BundleFusion with vs2013, but librealsense2 need vs2015 or higher. When I switch to vs2015, I have a compile problem same as https://github.com/niessner/mLib/issues/11 It is Have you met the same problem? If not, I will let you know when I successfully run with my D435 and see if ply blurry or not.

[flyisme]

Thanks for your sharing! Now I have a new problem. I compile BundleFusion with vs2013, but librealsense2 need vs2015 or higher. When I switch to vs2015, I have a compile problem same as https://github.com/niessner/mLib/issues/11 It is Have you met the same problem? If not, I will let you know when I successfully run with my D435 and see if ply blurry or not.

I did not encounter this problem, it may be that one of my dependent libraries is relatively old

[flyisme]

other question

getHeapFreeCount() It is always decreasing most of the time, and finally there is not enough memory，I just modified "SDF Voxel Size = 0.004f;"

[chethanab16]

@flyisme @tyronedong ,I am Trying to compile Bundle fusion with win10(cuda10.2 +vs 2017, but I am Getting the following Error ,

I also Tried using the new mlib But still got the error as follows ,Could some one give me the suggestion how to fix this issue Thankyou

I fixed this error _Severity Code Description Project File Line Suppression State Error (active) E1696 cannot open source file "boost/serialization/array_wrapper.hpp" FriedLiver D:\Azure_Kinect_BundleFusion\BundleFusion-master1_2017retar\external\mLib\include\core-base\common.h 61 But still have the following issue

[lombardm]

other question

getHeapFreeCount() It is always decreasing most of the time, and finally there is not enough memory，I just modified "SDF Voxel Size = 0.004f;"

Hi @flyisme , did you manage to solve this? I am encoutering the same problem. Everything works fine with default settings but when I try do decrease the voxel size it crashes after two hundreds of frames. I also tried to modify s_hashNumBuckets in zPrametersDefault but without success!

[chethanab16]

Hi Any one successfully compiled Bundle Fusion with VS 2017

[BruceWANGDi]

@flyisme Hi, have you figured out any solution to improve resolution? The .ply document is looked very blurry.

[chethanab16]

@BruceWANGDI thankyoufor providing me the link, yes as you mention array_wrappe is missing and in issue #57 they specifide it for VS 2013 which works fine but when i try to add the array_wrapper.hpp to the location \mLibExternal\include\boost\serialization then again I will receive the following error

On Mon, Oct 12, 2020 at 4:13 PM BruceWANGDi notifications@github.com wrote:

@chethanab16 https://github.com/chethanab16

57 https://github.com/niessner/BundleFusion/issues/57 is helpful

The reason why the error described on your second photo occurred is that you lose the document 'array_wrapper.hpp', you should put this document in your library(the location is \mLibExternal\include\boost\serialization), you can find 'array_wrapper' by searching its name on google.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/niessner/BundleFusion/issues/63#issuecomment-706923727, or unsubscribe https://github.com/notifications/unsubscribe-auth/AIDY4AYM6GSZMUNI7AGGXLLSKKUBHANCNFSM4PP24LIA .

[H-tr]

win10(cuda10.2 )+GeForce GTX 1060+vs 2017+ intelD415

Hi, have you figured out the reason? I'm facing the same problem using Realsense D435. I'm wondering if it is a problem with the setting or camera input quality is not good.