Hello everyone, I trained my own dataset on AutoML (google cloud), exequted it as “TF Lite” and ones as “Container”, and now I have to implement it to my code.
I saw examples with “cv2.dnn_DetectionModel” but where should I take this config file (.pbtxt)?
Attach screen of files that I have.
Will very appreciate for your help
P.S. may be the problem is that I have only one lable in the model?
to my knowledge, there is no current(4.5.2) support for tflite models
(only graphdef based tf 1.x models )
and the pbtxt files are optional, maybe you do not need one for your net
And .pb file? If I want to do some manipulation about the information I get from the webcam stream. Do you know some good tutorials I may use?
you mean the
this might still be a problem, you have to try …
what does your network do ?
Yes, I mean saved_model.pb. I build raspberry pi project. Right now it will be enough while the system recognizes the object it will light ON a LED.
again, try and report errors, please.
Are you using the readNet() from DNN module? You can place the model where it would be more convenient and pass the full path. From the docs, a .pb model should work and no tflite support is provided, at least until 4.5.2, as @berak already mentioned.
I got success to work with tflite models using the tensorflow/tflite directly. It is a bit harder, so stick with opencv if you can, otherwise it would solve your problem in a harder way. Tested on Python and C++.