3DScan Memory Issue

[KYOUNGRYOL_K_]

Hi, 

Nowadays I've been tested 3DScan module with RealSense F200 sensor which I have. The quality of the constructed model was pretty good so that I expect that we could be doing well in this 3D reconstruction domain (comparing with Kinfu by Microsoft).

BTW, it consume more than 450 MB of memory when I call the function reconstruct(*) - it's okay though - it does not release memory even I call Dispose() method and call GC.collect() in C# wrapper, but I think this occurs even in C++ library which I cannot access to the source code for the modification to release this unmanaged memory.

I, as an Intel Software Engineer, too, think this could be very critical issue. Hope this bug could be fixed asap. Thanks.

[MartyG]

Memory usage was being discussed recently on the post below.  It could be helpful to you.

https://software.intel.com/en-us/forums/topic/545104

[KYOUNGRYOL_K_]

Marty,
thanks for your comment.

Actually I'm pretty sure that the function reconstruct(*) in 3DScan module shows memory leak, which I don't think this is my programming issue, but the memory leak bug in the module.

[Timothy_T_]

I got the same problem too.....
this only happened after Reconstruct(*)...

Here is my code : 
result = my3DScan->Reconstruct(format, filename, reconstructionOptions);
it caused memory leak, but the memory did be released if I didn't do Reconstruct(*).

I think this is an Intel RSSDK bug.

Another issue is when I tried to use System::ComponentModel::backgroundWorker to implement DF_3DScan Sample.
The memory usage grows horribly when called function PreviewImageContainsObject(PXCImage* preview_image),
which caused my3DScan->SetMode(PXC3DScan::SCANNING) always equal PXC_STATUS_ITEM_UNAVAILABLE 
that I cannot do scan anymore unless I restart my program.

But this memory issue doesn't occur in the original single thread DF_3DScan sample code.

[Pierre_P_]

Hi,

I'm getting exactly the same memory leak in the C++ code as well.

[Colleen_C_Intel]

thanks for reporting this issue. It is being investigated by Intel.

[Pierre_P_]

Thanks Colleen. I posted earlier about this as well. By the way I'm getting a memory leak in Reconstruct() in both the V4 and V5 (R3) SDK versions, although the leak is significantly bigger in V4