Hi, I’m trying to use ScalableFusionCuda example in demo folder of https://github.com/theNded/Open3D.git and I use the data in the TestData folder.
But an error occurs:
CUDA error at /home/stereye/Open3D/src/Cuda/Integration/ScalableMeshVolumeCudaKernel.cuh:65 code=700(cudaErrorIllegalAddress) “cudaDeviceSynchronize()”
free():invalid size
My work environment:
Ubuntu18.04 RTX2070
Here is my code:
#include
#include#include <Open3D/Open3D.h>
#include <Cuda/Open3DCuda.h>#include “Utils.h”
using namespace open3d;
using namespace open3d::camera;
using namespace open3d::io;
using namespace open3d::geometry;
using namespace open3d::utility;int main(int argc, char *argv[]) {
using namespace open3d;std::string base_path = "/home/stereye/Open3D/examples/TestData/RGBD/"; auto camera_trajectory = CreatePinholeCameraTrajectoryFromFile(base_path + "trajectory.log"); auto rgbd_filenames = ReadDataAssociation(base_path + "rgbd.match"); int index = 0; cuda::PinholeCameraIntrinsicCuda intrinsics( PinholeCameraIntrinsicParameters::PrimeSenseDefault); float voxel_length = 0.01f; cuda::TransformCuda extrinsics = cuda::TransformCuda::Identity(); cuda::ScalableTSDFVolumeCuda tsdf_volume( 8, voxel_length, 3 * voxel_length, extrinsics); Image depth, color; cuda::RGBDImageCuda rgbd(640, 480, 4.0f, 5000.0f); cuda::ScalableMeshVolumeCuda mesher(cuda::VertexWithNormalAndColor, 8, 120000); std::shared_ptr<cuda::TriangleMeshCuda> mesh = std::make_shared<cuda::TriangleMeshCuda>(); Timer timer; for (int i = 0; i < rgbd_filenames.size() - 1; ++i) { LogDebug("Processing frame {} ...\n", index); ReadImage(base_path + rgbd_filenames[i].first, depth); ReadImage(base_path + rgbd_filenames[i].second, color); rgbd.Upload(depth, color); /* Use ground truth trajectory */ Eigen::Matrix4d extrinsic = camera_trajectory->parameters_[index].extrinsic_.inverse(); extrinsics.FromEigen(extrinsic); tsdf_volume.Integrate(rgbd, intrinsics, extrinsics); timer.Start(); mesher.MarchingCubes(tsdf_volume); timer.Stop(); *mesh = mesher.mesh(); index++; }}
For some unknown reason, I can’t use the visualizer, it will cause another error. Since I’m new to Cuda, it’s hard for me to debug.
Thank you a lot if someone could help me.