SeeNN: Leveraging Multimodal Deep Learning for In-Flight Long-Range Atmospheric Visibility Estimation in Aviation Safety
Taha Bouhsine, Giuseppina Carannant, Nidhal C. Bouaynaya, Soufiane Idbraim, Phuong Tran, Grant Morfit, Maggie Mayfield, Charles Cliff Johnson,
SeeNN is a multi-modal image classification system that leverages various image modalities to improve classification accuracy. This project is part of a research paper exploring advanced techniques in computer vision and deep learning.
- Multi-modal input support (RGB, Depth, Normal, Edge, Entropy)
- Flexible model architecture with attention-based fusion
- Customizable training parameters
- Integration with Weights & Biases for experiment tracking
- Comprehensive evaluation metrics and visualizations
- Python 3.7+
- TensorFlow 2.x
- CUDA-compatible GPU (recommended)
- Additional dependencies listed in
requirements.txt
To train the model:
bash train_all.shModify train_all.sh to adjust training parameters and enabled modalities.
For more control, use train_all.py directly.
See train_all.py for a full list of available arguments.
src/train_all.py: Main training scriptsrc/utils/: Utility functions for data loading, model building, etc.train_all.sh: Bash script for easy training execution
If you use this code in your research, please cite our paper: