Deploy to mobile android or convert to tf.lite format

Community Help
1

hi, i want to deploy my model to use in mobile android app, does anyone know how? or maybe how to convert it to tf.lite form?

2

If it is a Roboflow model, you can deploy it to a browser-based app that will work on Android:

We do not yet have a specific Android SDK for deployment.

For a custom Android-based app, I recommend following this guide to train a TFLite model:

  • Roboflow blog post
  • Here are the notebook options for training and creating your TFLite files, based on TensorFlow 2: 1 and 2

If you have any trouble, there are some troubleshooting tips and further instructions in the TensorFlow guide: Quickstart for Android  |  TensorFlow Lite

3

thanks mohamed for your reply, this really helped me finish my final project

4

when i try the notebook 2, i got a problem like this. can you help me to solve it? the first notebook works well with ver 8.0.56

ValueError Traceback (most recent call last)
in <cell line: 1>()
----> 1 model.export(export_dir=f’{dataset.location}', tflite_filename=tflite_filename,
2 label_filename=label_filename, vocab_filename=vocab_filename,
3 saved_model_filename=saved_model_filename,
4 export_format=export_format)#other formats: tfjs, saved_model, vocab, label

1 frames
/usr/local/lib/python3.9/dist-packages/tensorflow_examples/lite/model_maker/core/task/custom_model.py in _get_export_format(self, export_format, **kwargs)
89 for e_format in export_format:
90 if e_format not in self.ALLOWED_EXPORT_FORMAT:
—> 91 raise ValueError(‘Export format %s is not allowed.’ % e_format)
92
93 return export_format

ValueError: Export format tflite is not allowed.

5

You’ll just want to proceed with the first notebook in that case, since it is confirmed to be working for you - there may have been new changes to TF2 causing new bugs in “notebook 2.”

6

my final project is to compare the algorithm between efficientdet-lite and yolo v8. so i want to training with efficientdet-lite too

7

Hi there Mohamed, I will like to ask the model that deployed in browser-based app is still working in a flutter application or not?

8

Can you provide me with steps on how to do it?
I have trained a model on Roboflow, how do I integrate it with my android application?