Experimental Design for Bathymetry Editing

TL;DR

This paper explores the application of machine learning to bathymetry editing, revealing that real-world data often violate IID assumptions, which can negatively impact model performance when using standard data splitting methods.

Contribution

It demonstrates the limitations of IID assumptions in practical bathymetry editing tasks and highlights the need for alternative data splitting strategies.

Findings

Standard random data splits can lead to poor model performance.

Real-world bathymetry data often deviate from IID assumptions.

Machine learning applications must consider data distribution in practice.

Abstract

We describe an application of machine learning to a real-world computer assisted labeling task. Our experimental results expose significant deviations from the IID assumption commonly used in machine learning. These results suggest that the common random split of all data into training and testing can often lead to poor performance.