Hello,

I tried gSLICr by specifying different size of superpixel :

• For spixel_size = 10, it results in a strange segmentation like on the picture [slic_size10] , with totally squared spixel. Is it supposed to work only with specific spixel_size?

• For spixel_size = 50, there are still "orphan" superpixels as we can see on image [slic_size50]. Is the function "do_enforce_connectivity" not supposed to remove them?

  I have tried with other images and other machines but the problem seems recurrent.

This add a catkin/colcon metainfo file and installation instructions to the CMake file such that this repo can be used in a catkin/colcon workspace. It does not add any new dependencies.

Hey there,

I can't figure out how to access the superpixel result correctly. I don't need the drawn output image itself but the actual superpixel information. I basically want to store the superpixel ID for each pixel as a 4th channel in the image. Is there no trivial way to do this, or am I just missing something? I also need a way to determine neighboring superpixels, if it's possible.

Best, Alex

In the demo, why the following code gives me a segmentation image with three copies??

load_image(out_img, boundry_draw_frame);
		imshow("segmentation", boundry_draw_frame);

And the following code can't save the image to pgm. It only gives a black image as following, is there something wrong?

sprintf(out_name, "seg_%04i.pgm", save_count);	gSLICr_engine->Write_Seg_Res_To_PGM(out_name);

GitHub changed the way Markdown headings are parsed, so this change fixes it.

See bryant1410/readmesfix for more information.

Tackles bryant1410/readmesfix#1

gSLICr_core_engine.obj : error LNK2019: Unresolved external symbols "public: __cdecl gSLICr::engines::seg_engine_GPU::seg_engine_GPU(struct gSLICr::objects::settings const &)"

This is the same as https://github.com/carlren/gSLICr/pull/26 with a different branch name. The previous PR was closed automatically after the branch was renamed.

This adds a catkin/colcon metainfo file and installation instructions to the CMake file such that this repo can be used in a catkin/colcon workspace. It does not add any new dependencies.

Ubuntu 16.04 build fails with the following output

/usr/include/string.h: In function ‘void* __mempcpy_inline(void*, const void*, size_t)’: /usr/include/string.h:652:42: error: ‘memcpy’ was not declared in this scope return (char *) memcpy (__dest, __src, __n) + __n; ^ CMake Error at gSLICr_lib_generated_gSLICr_seg_engine_GPU.cu.o.cmake:266 (message): Error generating file /home/paperspace/gSLICr/build/CMakeFiles/gSLICr_lib.dir/gSLICr_Lib/engines/./gSLICr_lib_generated_gSLICr_seg_engine_GPU.cu.o

CMakeFiles/gSLICr_lib.dir/build.make:265: recipe for target 'CMakeFiles/gSLICr_lib.dir/gSLICr_Lib/engines/gSLICr_lib_generated_gSLICr_seg_engine_GPU.cu.o' failed make[2]: *** [CMakeFiles/gSLICr_lib.dir/gSLICr_Lib/engines/gSLICr_lib_generated_gSLICr_seg_engine_GPU.cu.o] Error 1 CMakeFiles/Makefile2:104: recipe for target 'CMakeFiles/gSLICr_lib.dir/all' failed make[1]: *** [CMakeFiles/gSLICr_lib.dir/all] Error 2 Makefile:83: recipe for target 'all' failed make: *** [all] Error 2

how can you retreive the super pixels as cv::contours? or some data structure similar, so that you can then further process each super pixel individually

Each SM splits its own blocks into Warps (currently with a maximum size of 32 threads).

All the threads in a warp executes concurrently on the resources of the SM ?

__syncthreads();
	if (should_add)
	{
		if (local_id < 128)
		{
			color_shared[local_id] += color_shared[local_id + 128];
			xy_shared[local_id] += xy_shared[local_id + 128];
			count_shared[local_id] += count_shared[local_id + 128];
		}
		__syncthreads();

		if (local_id < 64)
		{
			color_shared[local_id] += color_shared[local_id + 64];
			xy_shared[local_id] += xy_shared[local_id + 64];
			count_shared[local_id] += count_shared[local_id + 64];
		}
		__syncthreads();

		if (local_id < 32)
		{
			color_shared[local_id] += color_shared[local_id + 32];
			color_shared[local_id] += color_shared[local_id + 16];
			color_shared[local_id] += color_shared[local_id + 8];
			color_shared[local_id] += color_shared[local_id + 4];
			color_shared[local_id] += color_shared[local_id + 2];
			color_shared[local_id] += color_shared[local_id + 1];

			xy_shared[local_id] += xy_shared[local_id + 32];
			xy_shared[local_id] += xy_shared[local_id + 16];
			xy_shared[local_id] += xy_shared[local_id + 8];
			xy_shared[local_id] += xy_shared[local_id + 4];
			xy_shared[local_id] += xy_shared[local_id + 2];
			xy_shared[local_id] += xy_shared[local_id + 1];

			count_shared[local_id] += count_shared[local_id + 32];
			count_shared[local_id] += count_shared[local_id + 16];
			count_shared[local_id] += count_shared[local_id + 8];
			count_shared[local_id] += count_shared[local_id + 4];
			count_shared[local_id] += count_shared[local_id + 2];
			count_shared[local_id] += count_shared[local_id + 1];
		}
	}
__syncthreads();

__syncthreads();
		if (local_id < 32)
		{
			color_shared[local_id] += color_shared[local_id + 32];
			xy_shared[local_id] += xy_shared[local_id + 32];
			count_shared[local_id] += count_shared[local_id + 32];
		}
__syncthreads();
if (local_id < 16)
		{
			color_shared[local_id] += color_shared[local_id + 16];
			xy_shared[local_id] += xy_shared[local_id + 16];
			count_shared[local_id] += count_shared[local_id + 16];
		}
__syncthreads();

I am new to CUDA, and its version is 9.0 in my PC. My compilation process failed, and it says, "nvcc fatal : 32 bit compilation is only supported for Microsoft Visual Studio 2013 and earlier ", could you kindly tell me how to convert it to 64 bit version?