search

lutzroeder / netron

Visualizer for neural network, deep learning and machine learning models
https://netron.app
MIT License
27.76k stars 2.76k forks source link

xxd hexdump encoded in source files support #574

Closed hpssjellis closed 4 years ago

hpssjellis commented 4 years ago

Feature Request: TensorflowLite .h files

Background: TensorlfowLite .tflite files for mobile platforms can be converted into .h (C header) files for micro-controllers using the command

xxd -i model.tflite model.h

I assume there is a way to convert these files back from .h to .tflite , but I have not yet found that command. It would be useful to be able to view the Netron output directly from a TensorflowLite .h header file as these can be available from the web for micro-controllers without the corresponding .tflite file.

It would also be good to be able to compare the 2 types of TensorflowLite files with other models from TensorflowJS, Keras and Tensorflow.

lutzroeder commented 4 years ago

The general case would be any arbitrary source code file containing any potential number of encoded model files. Not a priority given the error conditions and special cases this may produce. It should be possible to write the byte[] to a file.