Pose Estimation of Buried Deep-Sea Objects using 3D Vision Deep Learning Models

TL;DR

This paper introduces BarrelNet, a deep learning model that estimates the pose and burial fraction of deep-sea barrels from point clouds, improving accuracy over traditional methods using synthetic training and ROV footage.

Contribution

The paper presents BarrelNet, a novel deep learning approach for 6-DOF pose estimation of buried seabed barrels from point clouds, trained on synthetic data and validated on real ROV footage.

Findings

BarrelNet outperforms traditional least squares fitting in accuracy.

Synthetic barrel point clouds effectively train the model for real-world data.

The approach demonstrates potential for deep-sea debris monitoring.

Abstract

We present an approach for pose and burial fraction estimation of debris field barrels found on the seabed in the Southern California San Pedro Basin. Our computational workflow leverages recent advances in foundation models for segmentation and a vision transformer-based approach to estimate the point cloud which defines the geometry of the barrel. We propose BarrelNet for estimating the 6-DOF pose and radius of buried barrels from the barrel point clouds as input. We train BarrelNet using synthetically generated barrel point clouds, and qualitatively demonstrate the potential of our approach using remotely operated vehicle (ROV) video footage of barrels found at a historic dump site. We compare our method to a traditional least squares fitting approach and show significant improvement according to our defined benchmarks.