This repository contains an implementation of NVIDIA PilotNet, a deep learning-based end-to-end autonomous driving system. The original PilotNet model was introduced by NVIDIA to learn steering commands directly from raw input images using a convolutional neural network (CNN). This implementation aims to replicate and enhance the system with modern deep learning frameworks and additional features.
- End-to-end learning for autonomous driving.
- CNN-based architecture inspired by NVIDIA PilotNet.
- Trained on real-world or simulated driving datasets.
- Utilizes PyTorch and TensorFlow for model training and inference.
- Supports real-time inference on NVIDIA Jetson devices.
- Includes data preprocessing and augmentation techniques.
- Integrated visualization tools for steering angle prediction.
Make sure you have the following dependencies installed:
pip install torch torchvision tensorflow numpy opencv-python matplotlib- Integrate sensor fusion with LiDAR and IMU.
- Improve generalization with diverse datasets.
- Optimize model for lower latency on edge devices.