Forums - qtimletflite plugin

6 posts / 0 new
Last post
qtimletflite plugin
pe.ostad
Join Date: 21 May 21
Posts: 14
Posted: Wed, 2021-09-29 03:06

Hi Everyone,

The idea is loading and executing tflite models using qtimletflite plugin. it  already works with the model provided in the object detection sample app. But I inserted many other tflite models with different configurations (preprocessing, delegates, post processing, ..), and had no luck executing them. Do we need to change the code for this plugin (or possibly qtioverlay plugin) based on our application? 

By the way, I used the tflite models provided for GstInference plugin.

Best Regards,

Peyman

  • Up0
  • Down0
Mayur
Join Date: 26 Feb 21
Posts: 16
Posted: Tue, 2021-10-05 06:42

Hi , 

You can use gstreamer commands before changing code. please refer following command & change the parameter according to your path 

gst-launch-1.0 -e qtiqmmfsrc name=qmmf ! capsfilter caps="video/x-raw,format=NV12,width=1280,height=720,framerate=30/1,camera=0" ! qtimletflite config="/data/misc/camera/mle_tflite.config" model="/data/misc/camera/detect.tflite" labels="/data/misc/camera/labelmap.txt" postprocessing="detection"  !  queue  ! qtioverlay ! waylandsink fullscreen=true async=true

You can change config , model & labelmap  parameter for your model. also refer qtimletflite for parameters. 

if above command works correctly (with your changes) then change the following line in code 

/* Modify the properties for qtimle-tflite */
    g_object_set (qtimletflite, "config", "/data/misc/camera/mle_tflite.config", 
        "model", "/data/misc/camera/detect.tflite", "labels", 
        "/data/misc/camera/labelmap.txt", "postprocessing", "detection", NULL);

Thanks & regards,

Mayur 

  • Up0
  • Down0
pe.ostad
Join Date: 21 May 21
Posts: 14
Posted: Wed, 2021-10-06 01:19

Hi Mayur,

Thank you for your response. I've already tested the models with gst-launch-1.0 . Like explained, I used different paramters for the qtimletflite plugin for the models. I changed almost every paramter of the plugin (the config file, preprocessing, postprocessing etc...) , but still no luck. It works with the config file, model , and labelmap which are already provided in the sample app, but not with new models.

I wanted to know if anyone has already tried this plugin for their own model/project and were successful, and I have to change the plugin code according to my project.

Best Regards,

Peyman

 

  • Up0
  • Down0
zmonteci
Join Date: 22 May 18
Posts: 16
Posted: Tue, 2022-02-08 11:45

Did you ever solved this issue? I had the same problem.

Zilmara

  • Up0
  • Down0
mbauroth
Join Date: 23 Mar 23
Posts: 4
Posted: Fri, 2023-04-21 11:01

Same question here!

  • Up0
  • Down0
luis.merayo
Join Date: 13 Feb 23
Posts: 6
Posted: Fri, 2023-07-21 15:39

Hi everyone,

We have used the qtimletflite plugin for running a TFLite pretrained model from Google Coral. You can check it out in our comprehensive Wiki guide. We also have step by step guides for installing the Neural Processing SDK and using the qtimlesnpe plugin as well.

You can find the Wiki guide at the following link: https://developer.ridgerun.com/wiki/index.php/Qualcomm_Robotics_RB5/AI_hardware_acceleration/TensorFlow/Example_pipeline

  • Up0
  • Down0
or Register

Opinions expressed in the content posted here are the personal opinions of the original authors, and do not necessarily reflect those of Qualcomm Incorporated or its subsidiaries (“Qualcomm”). The content is provided for informational purposes only and is not meant to be an endorsement or representation by Qualcomm or any other party. This site may also provide links or references to non-Qualcomm sites and resources. Qualcomm makes no representations, warranties, or other commitments whatsoever about any non-Qualcomm sites or third-party resources that may be referenced, accessible from, or linked to this site.