I am trying to work on yolov3 from samples of openvino. I created IR format data from model optimizer and plugged in to samples code.
But the code runs saying read all params and video and then exits out without any output or information.
Is there anything that i am missing ?
Also i tried debugging through visual studio, but it gives pop up saying inference_engine.dll is not installed properly. Hence i ran setupvars batch file , still then it did not work. Whereas in command line it works without any output.
Any ideas, please ?
C:\Users\Ignitarium\Documents\Intel\OpenVINO\inference_engine_samples_2015\intel64\Debug>object_detection_demo_yolov3_async.exe -i C:\Users\Ignitarium\Downloads\FroggerHighway_1.mp4 -m C:\Users\Ignitarium\Documents\tensorflow-yolo-v3\yolo_v3.xml -d GPU
API version ............ 1.4
Build .................. 17328
[ INFO ] Parsing input parameters
[ INFO ] Reading input
[ INFO:0] VIDEOIO: Enabled backends(6, sorted by priority): FFMPEG(1000); MSMF(990); DSHOW(980); VFW(970); CV_IMAGES(960); CV_MJPEG(950)
[ INFO ] Loading plugin
This is how i ran for conversion :
python C:\Intel\computer_vision_sdk_2018.4.420\deployment_tools\model_optimizer\mo_tf.py --input_model yolo_v3.pb --tensorflow_use_custom_operations_config yolo_v3_changed.json --batch 1
also have one from following :
python C:\Intel\computer_vision_sdk_2018.4.420\deployment_tools\model_optimizer\mo_tf.py --input_model yolo_v3.pb --tensorflow_use_custom_operations_config yolo_v3_changed.json --input_shape [1,416,416,3]
Which version of the openvino you are using? I am using R4, In this release, I have a file name "yolo_v3.json' but there is no file 'yolo_v3_changed.json', and after conversion when I ran inference I am getting very less FPS around 2-FPS output, can you have any idea about how to increase the FPS for this yolo model
I am struggling to get it running. I am using following :
OpenVino Version : R4 (2018.4.420)
and yolo_v3_changed.json is the same file as that of yolo_v3.json, as i just took one backup.
Can you please help me how did you ran this ?
Can you paste your outputs here, please.
Mine it does not give any output, hence i am not sure how i can check for FPS.
is this issue just for your YOLO? Can you run the other samples without any issues? Also what is your OS and CPU model?
If only YOLO has issues perhaps you could read more about model optimization and inference in the other threads
> OpenVino Version : R4 (2018.4.420)
You may also want to try again with R5 - if it was a setup issue could be solved by a re-install of the SDK.
> I am getting very less FPS around 2-FPS output, can you have any idea about how to increase the FPS for this yolo model
I think this is typical speeds we are seeing for the 608 size on both CPU/GPU path. Faster CPUs or GT4 GPUs will be a bit faster.
Maybe try tiny yolo or size 320.
As per your advice, i uninstalled and installed R5.
My Demo application runs , whereas yolov3 does not run even with -h option.
There is some relation to setupvars.bash , when ran on same cmd line it shows help of yolo otherwise it does not.
Also i am not able to debug from visual studio too.