MonoSOWA: Scalable monocular 3D Object detector Without human Annotations

TL;DR

MonoSOWA introduces a scalable, annotation-free monocular 3D object detection method that leverages a novel motion model, enabling effective training without human labels and outperforming prior approaches.

Contribution

It presents a new training approach for 3D detection from monocular images without human annotations, using a Local Object Motion Model and dataset aggregation techniques.

Findings

Outperforms previous methods on three datasets without human labels.

Approximately 700 times faster than prior work.

Effective as a pre-training tool for supervised learning.

Abstract

Inferring object 3D position and orientation from a single RGB camera is a foundational task in computer vision with many important applications. Traditionally, 3D object detection methods are trained in a fully-supervised setup, requiring LiDAR and vast amounts of human annotations, which are laborious, costly, and do not scale well with the ever-increasing amounts of data being captured. We present a novel method to train a 3D object detector from a single RGB camera without domain-specific human annotations, making orders of magnitude more data available for training. The method uses newly proposed Local Object Motion Model to disentangle object movement source between subsequent frames, is approximately 700 times faster than previous work and compensates camera focal length differences to aggregate multiple datasets. The method is evaluated on three public datasets, where despite using no human labels, it outperforms prior work by a significant margin. It also shows its versatility as a pre-training tool for fully-supervised training and shows that combining pseudo-labels from multiple datasets can achieve comparable accuracy to using human labels from a single dataset. The source code and model are available at https://github.com/jskvrna/MonoSOWA.