Hi ,
I am trying facial expression recognition from the tutorial in qualcomm website,
https://developer.qualcomm.com/software/qualcomm-neural-processing-sdk/l...
It is mentioned to download the source from ,
https://github.com/globaledgesoft/on-device-facial-experssion-recognitio...
And the project hierarchy mentioned in has data folder which has input folder, raw images folder, fer2013.csv,label-file.txt .
but unfortunately those folders and files are not present in the data directory .
Can anyone please help on this?
Hi Prabukumar,
in data folder of the project, you have to create the files & folder for further use.
input folder is for keeping your image files, which you wanted to test, Make sure that images should be a proper human face cropped image.
raw-images folder is required for generating the raw images from input image folder. Script provided in same learning resouces blog.
fer2013.csv , you can download from kaggle link provided in github repository of project.
label-file.txt contains names of all 7 facial expression classes in alphabetical order, make sure each classname is wriiten in newline.
Thanks & Regards,
Sahil Bandar
Hi Sahil,
I tried the dlc file provided within that in the gstreamer pipeline specified below with the contenet on label-file.txt as,
/data # cat label-file.txt
Angry
Disgust
Fear
Happy
Sad
Surprise
Neutral
gst-launch-1.0 -e qtiqmmfsrc ! video/x-raw, format=NV12, width=1280, height=720, framerate=30/1, camera=0 ! qtimlesnpe model=/data/fer.dlc labels=/data/label-file.txt postprocessing=classification ! qtioverlay ! qtivtransform rotate=1 ! waylandsink async=true sync=false fullscreen=true enable-last-sample=false
But I am getting an error from gstreamer mentioned below,
0:00:02.436757552 4126 0x5589692a00 INFO qtimle snpe_base.cc:411:InitFramework: InitFramework Exit
0:00:02.436788958 4126 0x5589692a00 ERROR qtimle ml_engine_impl.cc:467:Init: Init: Failed to init framework
0:00:02.436830468 4126 0x5589692a00 ERROR qtimlesnpe mle_snpe.cc:611:gst_mle_snpe_set_info:<mlesnpe0> MLE init failed.
0:00:02.436983854 4126 0x5589692a00 WARN basetransform gstbasetransform.c:1362:gst_base_transform_setcaps:<mlesnpe0> FAILED to configure incaps video/x-raw, format=(string)NV12, width=(int)1280, height=(int)720, framerate=(fraction)30/1, camera=(int)0 and outcaps video/x-raw, format=(string)NV12, width=(int)1280, height=(int)720, framerate=(fraction)30/1, camera=(int)0
0:00:02.437242135 4126 0x5589692a00 INFO basetransform gstbasetransform.c:1308:gst_base_transform_setcaps:<mlesnpe0> reuse caps
0:00:02.437595000 4126 0x5589692a00 INFO qtimle snpe_base.cc:51:ConfigureRuntime: DSP runtime selected
0:00:02.437635156 4126 0x5589692a00 INFO qtimle ml_engine_impl.cc:437:Init: Init: Enter
0:00:02.438235677 4126 0x5589692a00 INFO qtimle ml_engine_impl.cc:453:Init: Init: Loaded model from /data/fer.dlc
0:00:02.438347760 4126 0x5589692a00 INFO qtimle snpe_base.cc:392:InitFramework: InitFramework Enter
0:00:02.438430416 4126 0x5589692a00 INFO qtimle snpe_base.cc:394:InitFramework: SNPE version: 1.50.0.2622
0:00:02.438462239 4126 0x5589692a00 INFO qtimle snpe_base.cc:115:SetBuilderOptions: SetBuilderOptions: Enter
0:00:02.452182343 4126 0x5589692a00 INFO qtimle snpe_base.cc:137:SetBuilderOptions: SetBuilderOptions: Exit
0:00:02.452342447 4126 0x5589692a00 ERROR qtimle snpe_base.cc:388:PrintErrorStringAndExit: error_code=312; error_message=Undefined error. error_code=312; error_message=Undefined error. SNPE model format version detected: 2.5.0; error_component=SNPE; line_no=59; thread_id=548304552416; error_component=SNPE; line_no=262; thread_id=548654844320
Do you have any ideas on it ?
Hi Prabukumar,
I will suggest to go through this documents given below for classification example of using labels-file.txt with gstreamer pipeline.
https://developer.qualcomm.com/project/image-classification-qcs610-devel...
Please refer to the project given below for the same.
https://github.com/globaledgesoft/scene-classification-gstreamer
Hope this will resolve your problem.
Regards,
Sahil Bandar