Open Jeetkarsh opened 5 years ago
@Jeetkarsh , please try with OpenVINO-R3.
Software Ubuntu* 16.04 LTS Intel® Distributon of OpenVINO™ toolkit 2018 (R3)
@zahidulhaque Thanks for your reply although i was able to successfully build the files but now getting the following issue.
Can you please paste the screenshot of the error log.
++ Is it some issue with media SDK
i dont see the error logs or any screenshot. Can you point what issue you are seeing. Please paste the complete logs and at which step the issue is seen
@zahidulhaque Thanks for your reply although i was able to successfully build the files but now getting the following issue.
Its says error occurred while playing the ad!
@Jeetkarsh, Can you please provide the Processor and Graphic Card details.
Vendor ID: GenuineIntel CPU family: 6 Model: 94 Model name: Intel(R) Core(TM) i7-6820HQ CPU @ 2.70GHz Stepping: 3 CPU MHz: 2712.001 BogoMIPS: 5424.00 Graphic Card : Nvidia
Any Updates on this issue.
The application is validated on IEI Tank* AIoT Developer Kit which has Intel® HD Graphics 530. In Kiosk Analytics, hardware-accelerated implementation of the Intel Media SDK is provided which requires Intel Hardware and drivers.
ageGroup: 1 ageGroup: 1 ageGroup: 1 ageGroup: 1 ageGroup: 1 ageGroup: 1 InferenceEngine: API version ............ 1.2 Build .................. 13911 [ INFO ] Parsing input parameters [ INFO ] Loading plugin CPU
API version ............ 1.2
Build .................. lnx_20180510
Description ....... MKLDNNPlugin
[ INFO ] Loading network files for Face Detection [ INFO ] Batch size is set to 1 [ INFO ] Checking Face Detection inputs [ INFO ] Checking Face Detection outputs [ INFO ] Loading Face Detection model to the CPU plugin [ INFO ] Loading network files for AgeGender [ INFO ] Batch size is set to 16 for Age Gender [ INFO ] Checking Age Gender inputs [ INFO ] Checking Age Gender outputs [ INFO ] Age layer: age_conv3 [ INFO ] Gender layer: prob [ INFO ] Loading Age Gender model to the CPU plugin [ INFO ] Loading network files for Head Pose detection [ INFO ] Batch size is set to 16 for Head Pose Network [ INFO ] Checking Head Pose Network inputs [ INFO ] Checking Head Pose network outputs [ INFO ] Loading Head Pose model to the CPU plugin [ INFO ] Emotions Recognition DISABLED [ INFO ] Execution successful Loading the video ../data/demoPeople.mp4
Playing Add for Gender : M, Age Group : 2 Playing Ad: maleAd.h265
Unique visitors count : 0
[ERROR], sts=MFX_ERR_UNSUPPORTED(-3), Init, m_mfxSession.Init failed at /home/ubuntu/Downloads/kiosk-analytics/kiosk/src/pipeline_decode.cpp:296
[ERROR], sts=MFX_ERR_UNSUPPORTED(-3), media_sdk, Pipeline.Init failed at /home/ubuntu/Downloads/kiosk-analytics/kiosk/src/sample_decode.cpp:97 Error occurred while playing the ad!
This is the complete dump. The model is able to infer perfectly only issue is when ad need to be played. By any chance can we remove this add playing component from the complete pipeline.
In the source code, kiosk/src/main.cpp file, call to media_sdk() function is made which decodes the video passed to it as the parameter and plays it. If you comment the line number 345, i.e if you don't call the media_sdk() function, you will be able to run the application without actually playing the ad. In this case, you will still get the name of the ad depending on the demographics and the other data related to the demographics.
Thanks a lot for your input. I was able to run through the sample. Just a quick concern.
On the Grafana dashboard where it shows the people statistics (i.e No of People, No of Male and No of Female) practically doesn't it make sense to apply 'person-reidentification-retail-0076' model[although I am not aware of underlying change we need to make in this source code to account for this] to count each unique person once. Right now its taking 5 consecutive frames into consideration[ I guess] for counting and if the person is standing in front of the cam long enough he/she would be counted more then once which might give you the wrong statistics at the end of the day as to how many people look at your ad.