IMU Motion Analysis Toolkit

Tools for IMU sensor data processing and visualization

Usage

git clone --depth=1 https://github.com/FluxSand/Dataset.git Dataset
python train.py
# cp model.onnx XXX

📦 Files

1. CNN Motion Classifier (train.py)

Features:

• Advanced data preprocessing pipeline
  • Sequence padding & normalization
  • Stratified dataset splitting
• ONNX model export capability
• Embedded early stopping & learning rate scheduling

2. IMU Data Visualizer (imu_visualizer.py)

Features:

• Interactive CSV file management
  • File browsing & deletion
  • Auto-refresh directory
• Multi-view sensor visualization
  • 3D pose animation with adjustable parameters
  • Synchronized 2D plots (Pitch/Roll/Gyro/Accel)
  • Enhanced real-time data updates
• Customizable playback controls
  • Frame skip ratio (1-50)
  • Speed adjustment (10-200ms)
  • Interactive zoom and rotation

Requirements:

numpy~=2.0.2
pandas~=2.2.3
tensorflow~=2.18.0
tf2onnx~=1.16.1
onnx~=1.17.0
scipy~=1.15.1
imblearn~=0.0
scikit-learn~=1.6.1
matplotlib~=3.10.0
ttkbootstrap~=1.10.1

📂 Dataset Structure

Each CSV should contain:

Pitch,Roll,Gyro_X,Gyro_Y,Gyro_Z,Accel_X,Accel_Y,Accel_Z

📸 Sample Output

Classifier Training:

Epoch 150/2000
187/187 [==============================] - 15s 78ms/step - loss: 0.0327 - accuracy: 0.7215
Val accuracy: 0.6982 → Best model saved

Visualizer Interface: