adithya-samavedhi / e2e-tflite-tutorials

For project tracking of the Mobile ML WG: E2E TFLite tutorials and sample app ideas etc.

E2E TFLite Tutorials

Here is a list of sample app ideas and projects.

We would love your help! You can help by creating a tflite model ready for implementation, add a mobile app idea that needs a tflite model created, or write an end-to-end tutorial with sample code.

This is also where you can seek help from the community.

Project ideas (help needed!)

If you are interested in helping out, take a look at the potential projects below and assign an corresponding issue to yourself from the repo issue list.

Once you form a project team, move the idea to the "in progress" section below, create a new repo and link to it.

• YOLO - overview.
• Classify pose - overview.
• Sound classification - overview.
• Speech Command - overview.

Here are some more details on how exactly you can help:

• Generate ideas
• Create tflite model(s)
• Create Colab Notebook(s) demonstrating the model creation process along with running inference in Python
• Publish tflite model(s) optional
• Develop Android app(s) to demonstrate the model(s)
• Develop iOS app(s) to demonstrate the model(s)

The tutorials listed below would give you a good idea of the afore-mentioned pointers. Here you can find more detailed instructions on how to contribute.

End-to-end tutorials (in progress)

Take a look at the in progress projects to see what it's like to work on a project.

• Deeplab V3 - image segmentation. Colab Notebooks demonstrating the TFLite model conversion process for a variety of DeepLab V3 models along with running inference in Python are available on TF Hub. The TFLite variants of the DeepLab V3 models are also available from the same link.
• Mask-RCNN object detection, which is one of the most popular on-device ML use cases.
• DeepSpeech - a very popular ASR framework.
• Segmentation + Style Transfer - project repo.
• arbitrary_image_stylization by Magenta - art generation. Colab Notebooks demonstrating TFLite model conversion process along with inference in Python are available on TF Hub. The TFLite models can be downloaded from the same link.
• Enhanced super res GAN - project repo.
• CartoonGAN - project repo.

End-to-end tutorials (completed)

Once a project has been completed, please open a PR to awesome-tfite to add the links of the tflite model, sample code and tutorials.

• U-GAT-IT (Selfie <-> Anime) - project repo.
• SPICE (Pitch Detection) - Project repo - Medium article.

Contact the admins

This should be done via GitHub issues. Only use it when you have something relevant to discuss otherwise the issues will be automatically closed.