Data

Tools

Indexes

Applications

Experiments

Open Source Science

https://github.com/ahmetpala/clustering_ssl

https://github.com/ahmetpala/clustering_ssl

Science Score: 26.0%

This score indicates how likely this project is to be science-related based on various indicators:

  • ○
    CITATION.cff file
  • ✓
    codemeta.json file
    Found codemeta.json file
  • ✓
    .zenodo.json file
    Found .zenodo.json file
  • ○
    DOI references
  • ○
    Academic publication links
  • ○
    Academic email domains
  • ○
    Institutional organization owner
  • ○
    JOSS paper metadata
  • ○
    Scientific vocabulary similarity
    Low similarity (9.1%) to scientific vocabulary
Last synced: 10 months ago · JSON representation

Repository

Basic Info
  • Host: GitHub
  • Owner: ahmetpala
  • Language: Python
  • Default Branch: main
  • Size: 11.7 KB
Statistics
  • Stars: 1
  • Watchers: 1
  • Forks: 0
  • Open Issues: 0
  • Releases: 0
Created over 1 year ago · Last pushed 12 months ago
Metadata Files
Readme

README.md

Clustering-based analysis framework for underwater acoustic data

This repository contains code implementing a clustering-based framework for analyzing echograms, designed to identify sandeel regions using both self-supervised learning (SSL) features and raw data. Below is an overview of the key steps and functionality:

Features and Functionality

  1. Data Preprocessing:

    • Loads echogram data, labels, and bottom detection information from pre-defined directories.
    • Extracts multi-frequency data and applies transformations for standardization and filtering.

  2. Patch-Based Feature Extraction:

    • Divides echograms into patches (e.g., 8x8 pixels) for feature representation.
    • Extracts SSL-based features for each patch using the Extraction_for_Clustering module.

  3. Clustering:

    • Applies k-means clustering to the extracted features and raw data to identify clusters.
    • Generates cluster maps and channels for further analysis.

  4. Cluster Selection:

    • Implements an iterative cluster selection process to optimize metrics like the F1 score for identifying sandeel regions.
    • Evaluates individual clusters and their combinations for maximizing classification performance.

  5. Evaluation:

    • Calculates F1 score, precision, and recall for each cluster combination.
    • Compares the SSL-based clustering results with raw data clustering and supervised U-Net predictions.

  6. Visualization:

    • Provides visualization tools for comparing raw data, SSL-based clusters, U-Net predictions, and ground-truth labels.

Advantages

  • Handles class imbalance through over-clustering and iterative cluster selection without altering the original data distribution.
  • Offers flexibility by supporting multiple clustering methods, such as k-means, DBSCAN, and GMM.
  • Enables detailed comparative analysis with supervised models like U-Net, demonstrating the potential of clustering-based approaches for fisheries acoustics.

Applications

This framework is applicable to a range of fields, including: - Fisheries Acoustics: Identifying sandeel and other marine species.

For additional details and replication, refer to the corresponding author.

Owner

  • Login: ahmetpala
  • Kind: user

GitHub Events

Total
  • Watch event: 1
  • Push event: 4
  • Create event: 2
Last Year
  • Watch event: 1
  • Push event: 4
  • Create event: 2