Is it possible to use MacBook intel integrated GPU for object detection (Inference)?

[emregun]

Hi,

Problem:

I have been tried several object detection models and algorithms to optimise my object detection project(Basically continous object detection on several photos with 6 object classes in a loop), but unfortunately I am not so happy with my output since my CPU makes too much noise during object detection inference.

I guess the key point is using GPU, but I could not find a way to use my Intel integrated GPU for the object detection on my Macbook and when I use CPU only solutions, my project overloads and overheats CPU. In my activation monitor, %CPU value moves between %180-280.

My Requirements:
Silent CPU and Fan: I guess Activation Monitor %CPU should be maximum between %30-50
FPS: Min 0,5 fps(1 frame per 2 seconds) object detection.

What I have tried before:(just CPU only)
Tensorflow with faster_rcnn_inception_v2_coco(ROCm, XLA, AVX and AVX2 activated):
1 frame per 2 seconds, very good prediction but CPU too loud (%CPU:180-280) and overheated
Tensorflow faster_rcnn_inception_v2_coco(ROCm, XLA, AVX and AVX2 activated) with Openvino:
1 frame per 1,7 seconds, very good prediction but CPU too loud (%CPU:180-280) and overheated
Yolov3:
1 frame per 2,7 seconds, bad prediction and  CPU too loud (%CPU:120-190) and overheated
Yolov3 with openvino:
bad prediction and CPU and overheated
Yolov3-tiny:
1 frame per 0.9 seconds, horrible prediction and CPU loud (%CPU:70-80) and overheated

Environment:
MacBook Air (Retina, 13-inch, 2019)
macOS Catalina
1,6 GHz Dual-Core Intel Core i5
Intel UHD Graphics 617 1536 MB
Python3.6
Tensorflow 1.14.0

I would be grateful for any recommendations to use my Intel GPU for object detection or any other solutions.

Thanks a lot in advance

Emre

[Adli]

Hi Emre,

Thank you for reaching out to us. As for now, OpenVINO™ Toolkit is not supported for GPU on macOS. I apologize for the inconvenience.

Regards,

Adli

[Munesh_Intel]

Hi Emre,

This thread will no longer be monitored since we have provided a solution. If you need any additional information from Intel, please submit a new question.

Regards,

Munesh