Local

Env

OS Windows Subsystem Linux on Windows 11
Python v3.8.10
Pip v22.0.3

Install

pip install tensorflow==2.7.0
pip install -r requirements

Execute

python example.py -l labeled_anomalies.csv

NVIDIA Jetson Nano

Env

OS Ubuntu 18.04
Python v3.6.9
Pip v21.3.1
JetPack v4.5
CUDA v10.2
CUDNN v8.0.0

Install

CPU Version

pip install tensorflow -f https://tf.kmtea.eu/whl/stable.html

GPU Version

sudo pip3 install -U h5py==3.1.0
sudo pip3 install --pre --extra-index-url https://developer.download.nvidia.com/compute/redist/jp/v45 tensorflow
sudo pip3 install -r requirements

Execute

OPENBLAS_CORETYPE=ARMV8 python3 example.py -l labeled_anomalies.csv

Raspberry Pi3 (Model B)

Env

Architecture aarch64
OS Raspbian 11 (bullseye)
Python v3.9.2
Pip v20.3.4

Install

CPU Version [TF v2.7.0]
pip install tensorflow -f https://tf.kmtea.eu/whl/stable.html
or if needed a specify TF version
pip install tensorflow==2.7.0 -f https://tf.kmtea.eu/whl/stable.html
sudo pip install -r requirements

Known Problems

• ESN 2L: TFLite prediction killed at 0% cause 1.05GB of swap memory is not enough (Jetson has 2G of swap memory)