FSR Data Prediction

This is a neural network learning project aimed at obtaining accurate results without the need for expensive equipment when measuring the pressure applied to the soles of the feet in various situations. It allows obtaining results measured through F-Scan using 6 FSR sensors.

Ref: https://www.mdpi.com/1424-8220/18/12/4349

Structure

name	information
data	preprocessed data
eda	data eda
fsr_data	data loading process
fsr_model	deep learning model (implemented pytorch)
fsr_trainable	training process (implemented ray)
task*	hyper parameter tuning process

Tuning Result

https://wandb.ai/seokjin/FSR-prediction?workspace=user-seokjin

Result Summary

predict force RMSE

model	task1	task2	task4	task5
ANN	200.373	200.127	177.361	184.365
CNN-LSTM	203.214	178.705	171.708	175.383
LSTM	192.334	155.939	179.106	188.688

predict coord RMSE

model	task1	task3	task4	task6
ANN	0.943	0.6551	0.974	0.6716
CNN-LSTM	0.927	0.5428	1.022	0.6782
LSTM	0.901	0.5864	1.005	0.6762

simulation result

hyperparameter tuning result

Environment

Windows10 22H2

WSL Ubuntu 22.04.2

python 3.10.12

torch 2.0.1

ray[all] 2.5

Use Wandb

pip install wandb
wandb login

add ray.air.integrations.wandb.WandbLoggerCallback to ray.tune.Tuner

Use Memory Monitor

Setup Prometheus

recommended to download latest version

wget https://github.com/prometheus/prometheus/releases/download/v2.45.0/prometheus-2.45.0.linux-amd64.tar.gz
tar xvfz prometheus-2.45.0.linux-amd64.tar.gz
cd prometheus-2.45.0.linux-amd64
./prometheus --config.file=/tmp/ray/session_latest/metrics/prometheus/prometheus.yml

prometheus reference

ray reference

Setup Grafana

wget https://dl.grafana.com/enterprise/release/grafana-enterprise-10.0.1.linux-amd64.tar.gz
tar -zxvf grafana-enterprise-10.0.1.linux-amd64.tar.gz
cd grafana-10.0.1
./bin/grafana-server --config /tmp/ray/session_latest/metrics/grafana/grafana.ini web

grafana reference

ray reference

give authentication: here

if first grafana login, user name and password should be entered as 'admin' each.