README.md

Time Series Classification on Raspberry Pi using TensorFlow Lite

This example predicts a value using time series data from a Smart Factory.

1. Train model using Google Colab
2. Convert model to compressed version for Tensorflow Lite
3. Perform Inference on a Raspberry Pi using Tensorflow Lite

Dataset: https://www.kaggle.com/inIT-OWL/versatileproductionsystem

Overview:

Training and TFLite Compression

1. Open notebook from Google Colab: edge_time_series_classification.ipynb
2. Make a copy of the notebook
3. Run each cell of the notebook, step by step
4. Download *.pkl and *.tflite from Colab storage to your laptop

Inference

This phase uses the TensorFlow Lite Python API to perform inference using your compressed TF Lite model.

Documentation: https://www.tensorflow.org/lite/guide/inference#running_a_model

1. Deploy the *.pkl and *.tflite files to the Raspberry Pi. You can use WinSCP (on Windows) or scp (on MacOS) to transfer files to the Raspberry Pi. You should copy the files to this folder:

~/diec/day3/rpi

2. From the Raspberry Pi 3, launch docker image containing Tensorflow 2.0 for Python 3.7

cd ~/diec/day3/docker
sh ./launch_docker.sh

3. From the docker container running on the Raspberry Pi 3, evaluate the saved TF Lite model using inputs

cd day3/rpi
python3 evaluate.py input_file