- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi everyone,
I have been developing an application on Ubuntu 16.04, I use as a base the SqueezeNet cpp example on the ncsdk. (for the API calls)
when I compile on my laptop, it runs smoothly, no errors most of the times. The problem comes when i want to build the same example on the Pi.
I compile using:
g++ -std=c++11 app.cpp fp16.c -o app
pkg-config --cflags --libs opencv
-lmvnc
The problem is that i can't load the tensor
retCode = mvncLoadTensor(graphHandle, frame2, lenBufFp16, NULL);
if (retCode != MVNC_OK) //this gets me a -2
{ // error loading tensor
cout << "[ERROR]:Could not load tensor\n" << endl;
}
any advices??
I thought the graph was the problem but i could allocate it. I even run the same code in Python and it works, my question is why can I load the tensor on the Pi.
Link Copied
1 Reply
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
@longino Can you provide your app code (github) so we can look at it? It could be related to a image preprocessing issue or a fp32/fp16 conversion issue.
Reply
Topic Options
- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page