FSOCO: The Formula Student Objects in Context Dataset

TL;DR

FSOCO addresses the need for high-quality, domain-specific perception data for Formula Student Driverless cone detection. It introduces a collaborative, data buy-in dataset with bounding boxes and instance segmentation masks for cone classes, supported by labeling guidelines, automated QA, and tooling for data preparation. The paper analyzes dataset characteristics and demonstrates that models trained on FSOCO achieve better bounding-box regression performance than on the legacy dataset, despite fewer images, due to improved data quality and diversity. It also outlines a practical contribution workflow and plans to extend modalities and establish non-public test datasets to enable future competitions.

Abstract

This paper presents the FSOCO dataset, a collaborative dataset for vision-based cone detection systems in Formula Student Driverless competitions. It contains human annotated ground truth labels for both bounding boxes and instance-wise segmentation masks. The data buy-in philosophy of FSOCO asks student teams to contribute to the database first before being granted access ensuring continuous growth. By providing clear labeling guidelines and tools for a sophisticated raw image selection, new annotations are guaranteed to meet the desired quality. The effectiveness of the approach is shown by comparing prediction results of a network trained on FSOCO and its unregulated predecessor. The FSOCO dataset can be found at https://fsoco.github.io/fsoco-dataset/.

Figures (10)

• Figure 1: The FSOCO dataset contains human annotated bounding boxes and instance-wise segmentation masks for traffic cones as being used in the Formula Student Driverless competitions.
• Figure 2: FSOCO supports five object classes. The four main classes are shown here, the fifth class other includes all cones that are not rules compliant.
• Figure 3: Cones can be annotated by three different object tags, where (c) and (d) are combined in a single tag.
• Figure 4: Directory layout of the FSOCO dataset.
• Figure 5: Example images showing bounding boxes for different lighting and weather conditions.