Seamount prediction based on bathymetry data

Repository Link

https://github.com/annewoelfl/Seamounts

Description

• Less than 30 % of the seafloor has been mapped with high-resolution technology.
• Seamounts are of particular ecological and economic interest due to their unique ecosystems and mineral resources potential. They can also pose hazards to marine navigation and contribute to tsunami formation.
• Data from surveys comes piece by piece and is mostly processed manually and subjectively. Machine Learning could help to automate and speed up this process.

The models in this repository use images of bathymetry data and convolutional neural networks to predict if a seamount is on a provided input image.

Task Type

Computer Vision / Image Classification / Object Detection

Results Summary

• Best Model: cnn_model_1_0
• Evaluation Metric:[Accuracy, F1-Score]
• Result: [95% accuracy, F1-score of 0.94]

Documentation

1. Literature Review
2. Dataset Characteristics
3. Baseline Model
4. Model Definition and Evaluation
5. Presentation

Cover Image

Image generated with DALL-E 3 (based on a prompt requesting the similarity to Caspar David Friedrich: "Der Wanderer über dem Nebelmeer").