Device State Classification with Images and Dynamometer Data

Description

This is the repository for the ECE228 - Final Project by Team 24 (SP24).
It involves data classification with Mudestreda Multimodal Device State Recognition Dataset of Real Industrial Milling Device data containing Time Series and Image data.

Installation

To use this project, follow these steps:

1. Clone the repository to your local machine:

   git clone https://github.com/186shades/ECE228-Final-Project.git
   

2. Navigate to the project directory:

   cd ECE228-Final-Project
   

3. Run the ipynb files in following order:

   spec.ipynb (uncomment the first code block for downloading the dataset)
   tool.ipynb
   multi_recurrent.ipynb
   

---

Code Structure

Project Jupyter Notebooks:

• spec.ipynb has the model definitions and training code for the time series modality pathway
• tool.ipynb has the model definitions and training code for the image modality pathway
• multi_recurrent.ipynb has the final tempmixer block with GRU for the multimodal architecture along with it's training code

---

File Structure

1. output:
   • Contains the generated model checkpoints along with loss and accuracy across epochs.

---

Required Packages

torch

pandas

numpy

matplotlib

cv2

time

---

Team Members

• Eric Bressinger
• Nikhil Gandudi Suresh
• Sharvari Satish Deshmukh