Awesome

On-device Learning of Activity Recognition Networks

Personalized machine learning on the smartphone

The aim of the project is to provide an end-to-end solution for on-device training, inference and data collection for activity recongition based on <a href="https://github.com/tensorflow/examples/tree/master/lite/examples/model_personalization">TFlite Transfer Learning Pipeline</a>. The corresponding blog post is available <a href="https://aqibsaeed.github.io/on-device-activity-recognition">here</a>.

Tools Required

• Python 3.5+
• Tensorflow 2.0.0rc0
• Numpy
• Pillow
• Scipy
• Android Studio

Contribute (Future Work)

• Add support for pairing the app with a smartwatch and fine-tuning the model for a wearable device.
• Port SoundNet, add functionality for audio recording and tflite model conversion for handling dynamic size input.

If you are interested in contributing to this project, please submit a pull request or reach out at: aqibsaeed@protonmail.com.

Dataset

The <a href="https://archive.ics.uci.edu/ml/datasets/Heterogeneity+Activity+Recognition">Heterogeneity Activity Recognition dataset</a> is used for model pretraining. If you use this in your research, please cite their work and check the license.

Citing

If you find this project usefuly, please cite it as: