Application Acceleration With FPGAs
Programmable Acceleration Cards (PACs), DCP, FPGA AI Suite, Software Stack, and Reference Designs
477 Discussions

CL_OUT_OF_HOST_MEMORY error when running in HETERO mode (-d HETERO: FPGA, CPU).

SGrif4
Beginner
896 Views

Hardware/Software Used

Intel® Distribution of OpenVINO™ Toolkit 2018 R5

Intel® Vision Accelerator Design with Intel® Arria® 10 FPGA (IEI's Mustang-F100-A10)

 

 

Run pedestrian demo will mall.mp4 video following video tutorial (https://www.youtube.com/watch?v=OKH57mvO9k0),

but got errors.

 Takeaways:

Seems to run on CPU, will not run on FPGA due to “OUT OF HOST MEMORY” error, shown below.

 

Questions/Concerns:

·     Not really sure what the performance metric would be on this kind of demo / I don’t know where to output an FPS or other metric

o  Copy from face detection to get fps on pedestrian tracker

·     Works on CPU, does not run on HETERO:FPGA,CPU (error screenshot below)

 

Trying Different Bitstreams

 

·     MobileNet / SqueezeNet / Generic / AlexNet_GoogleNet / ResNet / TinyYolo_SSD300 / VGG error

error1.png

 

 

 

·     ELU / RMNet error

error2.png

 

o  Video pops up briefly then goes away, doesn’t show tracking mechanism

 

 

Notes / Appendix:

 

Code to run, from Release directory: ieisw@ieiswSER0:~/inference_engine_samples/intel64/Release$

 

./pedestrian_tracker_demo -i ~/FPGA_POV1/mall.mp4 -m_det /opt/intel/computer_vision_sdk/deployment_tools/intel_models/person-detection-retail-0013/FP32/person-detection-retail-0013.xml -d_det CPU -m_reid /opt/intel/computer_vision_sdk/deployment_tools/intel_models/person-reidentification-retail-0031/FP32/person-reidentification-retail-0031.xml -d_reid CPU -pc

 

./pedestrian_tracker_demo -i ~/FPGA_POV1/mall.mp4 -m_det /opt/intel/computer_vision_sdk/deployment_tools/intel_models/person-detection-retail-0013/FP32/person-detection-retail-0013.xml -d_det HETERO:FPGA,CPU -m_reid /opt/intel/computer_vision_sdk/deployment_tools/intel_models/person-reidentification-retail-0031/FP32/person-reidentification-retail-0031.xml -d_reid HETERO:FPGA,CPU -pc

0 Kudos
3 Replies
JohnT_Intel
Employee
610 Views

Hi,

 

May I know which board and OpenVINO version are you using? If you are using OpenVINO 2018 then could you upgrade it to OpenVINO 2019R1?

0 Kudos
SGrif4
Beginner
610 Views

Hardware/Software Used

Intel® Distribution of OpenVINO™ Toolkit 2018 R5

Intel® Vision Accelerator Design with Intel® Arria® 10 FPGA (IEI's Mustang-F100-A10)

 

Whoops, forgot to put that at the top. Thanks!

 

https://forums.intel.com/s/question/0D70P000006QkKSSA0

0 Kudos
JohnT_Intel
Employee
610 Views

Could you upgrade our OpeVINO to 2019R1.1 version?

0 Kudos
Reply