Transfer Learning with back propagation with tfilte model

google-coral / libcoral

C++ API for ML inferencing and transfer-learning on Coral devices

https://coral.ai

Apache License 2.0

83 stars 44 forks source link

Transfer Learning with back propagation with tfilte model #18

Open abucka opened 2 years ago

abucka commented 2 years ago

Description

Is it possible to also retrain a pure tflite model without edgetpu using back propagation method?

Click to expand!

### Issue Type Support ### Operating System Linux ### Coral Device _No response_ ### Other Devices _No response_ ### Programming Language C++ ### Relevant Log Output _No response_

hjonnala commented 2 years ago

Hello @abucka I think, it is possible. Please feel free to try to retrain pure tflite models from test_data repo using back propagation method.

abucka commented 2 years ago

Hi, I tested it but unfortunately the training process hangs and does not output any information about loss or accuracy like it happens in case of edgetpu version.

hjonnala commented 2 years ago

The process might be hanging due to CPU limitation. Can you try with one or two images from each dateset?