Pre-trained Models for Sonar Images

TL;DR

This paper introduces a new sonar image dataset and pre-trained neural networks to improve sonar perception, enabling effective low-shot classification and transfer learning across different sonar sensors.

Contribution

It provides the first publicly available pre-trained models and a dataset for sonar images, filling a significant gap in sonar machine learning resources.

Findings

Pre-trained models enable good classification with 10-30 samples per class.

Features transfer effectively across different sonar sensors.

Models trained on the dataset generalize well to other sonar datasets.

Abstract

Machine learning and neural networks are now ubiquitous in sonar perception, but it lags behind the computer vision field due to the lack of data and pre-trained models specifically for sonar images. In this paper we present the Marine Debris Turntable dataset and produce pre-trained neural networks trained on this dataset, meant to fill the gap of missing pre-trained models for sonar images. We train Resnet 20, MobileNets, DenseNet121, SqueezeNet, MiniXception, and an Autoencoder, over several input image sizes, from 32 x 32 to 96 x 96, on the Marine Debris turntable dataset. We evaluate these models using transfer learning for low-shot classification in the Marine Debris Watertank and another dataset captured using a Gemini 720i sonar. Our results show that in both datasets the pre-trained models produce good features that allow good classification accuracy with low samples (10-30 samples per class). The Gemini dataset validates that the features transfer to other kinds of sonar sensors. We expect that the community benefits from the public release of our pre-trained models and the turntable dataset.