Self-Supervised Person Detection in 2D Range Data using a Calibrated Camera

TL;DR

This paper introduces a self-supervised approach that leverages camera-based detectors to generate training labels for LiDAR-based person detection, reducing the need for manual annotations and improving performance across environments.

Contribution

The method automatically creates training labels from camera detections to enhance LiDAR-based person detectors without manual labeling, enabling better deployment adaptability.

Findings

Self-supervised detectors outperform those trained on different datasets.

Performance approaches that of manually annotated training.

Method reduces labeling effort and improves robustness.

Abstract

Deep learning is the essential building block of state-of-the-art person detectors in 2D range data. However, only a few annotated datasets are available for training and testing these deep networks, potentially limiting their performance when deployed in new environments or with different LiDAR models. We propose a method, which uses bounding boxes from an image-based detector (e.g. Faster R-CNN) on a calibrated camera to automatically generate training labels (called pseudo-labels) for 2D LiDAR-based person detectors. Through experiments on the JackRabbot dataset with two detector models, DROW3 and DR-SPAAM, we show that self-supervised detectors, trained or fine-tuned with pseudo-labels, outperform detectors trained only on a different dataset. Combined with robust training techniques, the self-supervised detectors reach a performance close to the ones trained using manual annotations of the target dataset. Our method is an effective way to improve person detectors during deployment without any additional labeling effort, and we release our source code to support relevant robotic applications.