meiqua/pose_refine

cuda icp for 6D pose estimation

pose_refine's Issues

Wrong Result When Render With ROI

First of all, thank you for your code!

I have tried the renderer_test (test.cpp) changing the ROI as follows :
cuda_renderer::Model::ROI roi = {320, 240, 100, 100};

And it give me the following result

When it should be:

Any clue why it is not working properly?

A second point I would like to raise relates about timing. The GPU ICP performance it's very impressive, good job! however this code:

device_vector_holder<::Vec3f> pcd_buffer_cuda, normal_buffer_cuda; scene.init_Scene_projective_cuda(scene_depth, K_, pcd_buffer_cuda, normal_buffer_cuda);

is about ten times slower than the ICP, why is it so slow? do you have any advice for improving its perfomance? Obviously, using a ROI will help and that's why it is so important that works properly the render with a ROI.

Thank you!

cuda_render test fail when each pose is different?

When I change the code here

std::vector<cuda_renderer::Model::mat4x4> mat4_v(100, mat4);
for (auto &mat:mat4_v)
 {
        mat.c3 += static_cast <float> (rand()) / static_cast <float> (RAND_MAX) * 100;
 }

Then the result of GPU and CPU is different. Is there anything I'm doing wrong?

Multiple ICPs

Hi,

I wanted to understand how to you use the code to do multiple ICPs at once. I tried to look at test.cpp but that seems to be doing one ICP if I am not wrong

kdtree building does not stop

When I use the code to building kdtree, it enters an endless loop, and there are always new nodes that can be split.

Got wrong Result

Hi.
I used your code to implement my own data.
I use PCL to read two point cloud from PCD files,
one is
Scene_nn scene;
KDTree_cpu kdtree_cpu;
scene.init_Scene_nn_cpu();

I also find normals from PCL and init KDTree in cup and push to GPU as init_Scene_nn_cuda does.

The other one is used by thrust::copy to get a device vector.

However, when I use cuda_icp::ICP_P2P_cuda to get results,
I find the first iteration works fine, but it gives a crazy extrinsic (variable in ICP_P2P_cuda funtion). Thus, the second iteration cannot find a reasonable results and Ab[28] is 0 to return the function. The final result is the crazy extrinsic in first iteration.

I am not sure why it give me this results and I use your data works very good.

It is appreciate that you could give me some hints.

Thanks in advance.

编译提示opencv头文件找不到

你好，我再windows下编译这个工程的时候，提示opencv头文件找不到，我可以确定路径设置正确。右键点对应头文件点打开文档也没有问题，就是编译提示找不到。

编译问题

你好，我在arm-linux编译的实例，提示要装open3d;我试着在linux下搭建open3d环境，sudo make -j6时，好多第三方库都clone下载不了；但是我copy连接，输入IE的URL，又可以下载。直接下载下来的源码放了一些位置，进行编译，还是不行。有好的方法推荐吗？谢谢。

about cuda_build_kdtree

Hi:
I follow the branch cuda_build_kdtree .I find there is a bug at pcd_scene.cu line 171 num_nodes_now_host = num_nodes_now_dev;

I change it to 'thrust::copy(num_nodes_now_dev.begin(), num_nodes_now_dev.end(), num_nodes_now_host.begin());' . But it have a bug, too.

Did you test this code?

interrupt error

When I run function "depth2cloud_cuda",it will interrupt in file"icp.cu " line 274.

Question about your CUDA render part

Dear author, this repo is really cool! I have a question about your CUDA render part.
I saw the function "render_triangle" seems to render tri-mesh. So:

1. Is the output Z-buffer of the rendered image?
2. In readme, you mentioned it can run "render handreds poes of a ply model at once". Is it possible to extend the code to render multiple object in each scene, and render all scenes at same time?
   Thank you for such amazing work!