Object Detection using Tensorflow Lite C API on Windows

This is the second video in the Crossplatform Tensorflow Lite series.
In this video we are developing our object detection module in C++ using Tensorflow Lite C API, and develop a test application in Windows to use our object detector.

[laurick]

ça faisait deux semaines que j'étais bloqué ; Merci pour ta vidéo et tes explications!

[rahulravindra]

Hello thanks a lot for the tutorial. Could you direct me to an example for tensorflow lite where it is purely run on C and not C++ ? Thank you

[tamilarvalvu]

Thanks a lot for the video tutorial it helps me a lot !! BTW which model are you using for object detection, is it Faster rcnn or a YOLO model ??

[陈周-zs]

thank you

The SSD model succeeded

But fast RCNN model failed

Can you tell me the default save from the training of fast RCNN model_ How does a model look like SSD save_ Model into a single file Lite or independent Pb can be used by other platforms

[陈周-zs]

But SSD training doesn't come out
Loss 0.3 doesn't go down
The fast RCNN loss can reach 0.002

[陈周-zs]

I mean, fast_ RCNN model
How to change the default
save_ model
Converted to a separate Pb file, it can be deployed to other platforms
Or a model that C + + can call (tensorflow2.4.1)

[gjergjishkurti]

Hey thanks a lot for the amazing tutorial!

I have a question regarding the input/output tensors
when i try to access members of the m_input_tensor i get an error saying "pointer to incomplete class type "TfLiteTensor" is not allowed"

For instance in the following snippet:

if (this->m_modelQuantized && m_input_tensor->type != kTfLiteUInt8) { // If quantized and not an Int
printf("Model is quantized but the input type is not an 8 bit int");
}

Have you had this issue b4, do you happen to know how it can be avoided/dealt with?
Thanks a lot again in advance! :)

[陈周-zs]

hello !
here error
// Create the interpreter.
m_interpreter = TfLiteInterpreterCreate(m_model, options);

ERROR: Didn't find op for builtin opcode 'CONV_2D' version '5'

ERROR: Registration failed.
Failed to create interpreter
How to deal with it?

[ooicheekean]

Hi, can this object detection deploy into microcontroller?