Visual-Inertial Odometry (VIO) Fusion with LSTM

This project implements a deep learning-based Visual-Inertial Odometry (VIO) system using PyTorch. It includes three model types: Vision-only, Inertial-only, and Fusion-based models, each using LSTM networks to estimate poses from simulated or real sensor data.

🧠 Model Types

1. VisionOnlyModel: LSTM model processing vision features.
2. InertialOnlyModel: LSTM model processing IMU sequences.
3. FusionModel: Combines both vision and inertial inputs through dedicated LSTMs and concatenates their outputs for pose estimation.

📁 File Structure

• vio.py – Core model definitions for VIO using PyTorch.

🛠️ Dependencies

• Python 3.8+
• PyTorch
• NumPy

Install dependencies using:

pip install torch numpy