https://github.com/sadmansakib93/arctic-mammal-classification

Deep learning model for acoustic detection and classification of Bowhead whales. This repo contains code for the binary classification model, two classes are: Bowhead whales (BH), Other/background.
https://github.com/sadmansakib93/arctic-mammal-classification

annotations audio-analysis audio-classification classification deep-learning deep-neural-networks machine-learning python spectrogram tensorflow

Last synced: about 1 month ago
JSON representation

Deep learning model for acoustic detection and classification of Bowhead whales. This repo contains code for the binary classification model, two classes are: Bowhead whales (BH), Other/background.

• Host: GitHub
• URL: https://github.com/sadmansakib93/arctic-mammal-classification
• Owner: SadmanSakib93
• Created: 2022-08-12T19:02:58.000Z (over 3 years ago)
• Default Branch: main
• Last Pushed: 2022-08-12T19:20:59.000Z (over 3 years ago)
• Last Synced: 2025-03-01T11:27:12.620Z (9 months ago)
• Topics: annotations, audio-analysis, audio-classification, classification, deep-learning, deep-neural-networks, machine-learning, python, spectrogram, tensorflow
• Language: Jupyter Notebook
• Homepage:
• Size: 2.93 MB
• Stars: 0
• Watchers: 3
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

          
# Arctic mammal classification

Deep learning model for acoustic detection and classification of Bowhead whales. This repo contains code for the binary classification model, two classes are: Bowhead whales (BH), Other/background.

Following are some statistics about the dataset and model:

- CB-50 and CB-300 dataset for training and testing

- CB-50 contains 11138 BH annotations, CB-300 contains 5983 BH annotations

- Average duration of BH annotations: 1.78 sec

- Average low and high frequency of BH: 90.38 - 530.80 Hz

- Spectrogram window size: 3 sec

- DenseNet model

## Training data distributation

![Alt text](images/train_distribution.png?raw=true)

## Overall workflow diagram

![Alt text](images/workflow.png?raw=true)

## Results

### Training & validation performance

Following image shows the training and validation loss curves

![Alt text](images/training_loss.png?raw=true)

### Test performance

Following graph shows the performance metrices on the test annotations

![Alt text](images/test_performance.png?raw=true)