Soft Labels for Rapid Satellite Object Detection

TL;DR

This paper explores the use of soft labels generated by models to rapidly create and enhance satellite object detection datasets, demonstrating that models trained on soft labels achieve near-original accuracy.

Contribution

It introduces a method to generate soft labels from model detections to quickly augment satellite object detection datasets, reducing annotation effort.

Findings

Soft labels enable training models nearly as accurate as those trained on original data.

Using soft labels accelerates dataset creation and improves model performance.

The approach is validated on the xView satellite dataset with YOLOv5.

Abstract

Soft labels in image classification are vector representations of an image's true classification. In this paper, we investigate soft labels in the context of satellite object detection. We propose using detections as the basis for a new dataset of soft labels. Much of the effort in creating a high-quality model is gathering and annotating the training data. If we could use a model to generate a dataset for us, we could not only rapidly create datasets, but also supplement existing open-source datasets. Using a subset of the xView dataset, we train a YOLOv5 model to detect cars, planes, and ships. We then use that model to generate soft labels for the second training set which we then train and compare to the original model. We show that soft labels can be used to train a model that is almost as accurate as a model trained on the original data.