Histogram-based Parameter-efficient Tuning for Passive and Active Sonar Classification

Amirmohammad Mohammadi, Davelle Carreiro, Alexandra Van Dine and Joshua Peeples

If this code is used, please cite it. (2025, March): Initial Release (Version v1.0).

arXiv

Zenodo archive

Zenodo. https://zenodo.org/records/15263760

Installation Prerequisites

The requirements.txt file includes all the necessary packages, and the packages will be installed using:

pip install -r requirements.txt

Or, for a more convenient option, a pre-configured environment is available for download through the provided link. This environment comes with all the essential packages already installed.

Download environment

Demo

To get started, please follow the instructions in the Datasets folder to download the DeepShip dataset. Next, run demo.py in Python IDE (e.g., Spyder) or command line to train, validate, and test models.

Inventory

https://github.com/Peeples-Lab/HLAST_DeepShip_ParameterEfficient 

└── root directory
    ├── demo_light.py                     // Main demo file.
    ├── Demo_Parameters.py                // Parameter file for the demo.
    ├── plot_curves.py                    // Run this after the demo to view learning curves. 
    ├── feature_similarity_analysis.py    // Run this after the demo to view feature similarites, PLEASE set the parameters accordingly. 
    └── Datasets                
        ├── Get_Preprocessed_Data.py       // Generate segments for the DeepShip dataset.
        └── SSDataModule.py                // Data Module for the DeepShip dataset.
        ├── ShipsEar_Data_Preprocessing.py // Generate segments for the ShipsEar dataset.
        └── ShipsEar_dataloader.py         // Data Module for the ShipsEar dataset.
        ├── Create_Combined_VTUAD.py 	   // Merge the three distinct scenarios into one for the VTUAD dataset.
        └── VTUAD_DataModule.py            // Data Module for the VTUAD dataset.
        └── fls_datamodule.py              // FLS (forward-looking sonar) vision DataModule.
    └── Utils                     
        ├── LitModel.py                    // Lightning Module for the the model.
        ├── Network_functions.py           // Contains functions to initialize the model.
        ├── LogMelFilterBank.py            // Log Mel Filter Bank Feature.
        └── Feature_Extraction_Layer.py    // Extract and transform features from the audio files.
    └── src
    	└── models              
		├── ast_base.py            // AST Original Model
		├── ast_linear_probe.py    // AST Linear Probing
		├── ast_adapter.py         // AST with Adapter Layers
		├── RBFHistogramPooling.py // Create the Histogram Layer
		└── ast_histogram.py       // AST with Histogram Layers (HPT)

License

This source code is licensed under the license found in the LICENSE file in the root directory of this source tree.

This product is Copyright (c) 2025 A. Mohammadi, D. Carreiro, A. Dine and J. Peeples. All rights reserved.

Citation

If you use this work, please cite:

arXiv preprint

@article{amir2025histogram,
  title={Histogram-based Parameter-efficient Tuning for Passive Sonar Classification},
  author={Mohammadi, Amirmohammad and Carreiro, Davelle and Van Dine, Alexandra and Peeples, Joshua},
  journal={arXiv preprint arXiv:2504.15214},
  year={2025}
}