computer vision with non-nvidia hardware

Question

My problem start when I was training my SSD model with an MXNet framework as here:https://github.com/zhreshold/mxnet-ssd I had this error:

File "/usr/lib/python2.7/ctypes/__init__.py", line 362, in __init__ 
self._handle = _dlopen(self._name, mode) OSError: libcudart.so.8.0: cannot 
open shared object file: No such file or directory

Searching for a solution leads me to install "Cuda" which I figure out later that it's just supported by NVidia graphics card. until now, I am not sure of my conclusion and how true is it, I don't have an NVidia card my computer is very simple Dell -core i3 with Intel(R)HD card.

In case this is true, what shall I do to run my code? I found this alternative solution: GPU Emulator for CUDA programming without the hardware and I've also read this: Can I run CUDA on Intel's integrated graphics processor?
If it's not related to that, what is the problem?

Sorry for my limit knowledge, I will appreciate any suggestions and descriptions.

Looking forward to hearing answers.

Regards.

score 0 · Answer 1 · answered May 25 '18 at 00:45

docker pull daviddocker78:mxnet-ssd:gpu_0.12.0_cuda9

The author is using a gpu-docker for their deployment. So you need a nvidia card. In my experience, trying to emulate hardware instead of buying a it is always lead to bitter pain. You time is limit so STOP what you are trying is better choice.

I guess you're trying some object-detection. Then why you don't try yolo yolo. It is state-of-the-art object detection and above of that it is very fast. Even on CPU you could achieve a 2-3fps.

computer vision with non-nvidia hardware

1 Answers1