PandaSet: Advanced Sensor Suite Dataset for Autonomous Driving
DOI: 10.1109/itsc48978.2021.9565009
archive: archived pipeline: cataloged verified
Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)
Summary
This paper introduces PandaSet, a multimodal dataset designed to address the critical need for high-quality, diverse data in autonomous driving research. The authors argue that current datasets often lack the sensor precision, annotation completeness, or environmental diversity required to train robust perception algorithms for Level 4 and 5 autonomy. Motivated by a reduction in road testing data during the COVID-19 pandemic, PandaSet is presented as the first open-source dataset featuring a complete, high-precision sensor suite with a no-cost commercial license. It aims to provide researchers with a resource that captures complex real-world scenarios, including varied lighting, traffic conditions, and challenging terrain. The dataset was collected using a Chrysler Pacifica minivan equipped with a comprehensive sensor array: one 360-degree mechanical spinning LiDAR (Hesai Pandar64, 200m range), one forward-facing long-range LiDAR (Hesai PandarGT, 300m range), six cameras (five wide-angle, one long-focus), and a GNSS/IMU unit. Data was gathered from two routes in Silicon Valley, resulting in 103 scenes, each lasting 8 seconds. The collection process utilized precise time synchronization via a trigger board and GPS clock, alongside rigorous sensor calibration to ensure alignment between LiDAR scans and camera images. Annotations were generated using multi-sensor fusion technology to achieve high precision, providing 3D bounding boxes for 28 object classes and semantic segmentation labels for 37 categories. PandaSet demonstrates significant advantages in data density and range compared to existing benchmarks like KITTI, nuScenes, and Waymo Open. The dataset captures objects up to 300 meters away, far exceeding the typical 70–75 meter limits of other datasets. Statistical analysis reveals a high density of labeled traffic participants; for instance, PandaSet contains an average of 133.5 cars per frame, compared to 26.5 in the Waymo Open dataset. The dataset covers diverse conditions, including night driving, construction zones, and hills, and includes rare object classes such as motorized scooters and animals. The authors provide baseline models for LiDAR-only 3D object detection, LiDAR-camera fusion 3D object detection, and LiDAR point cloud segmentation, along with a development kit to facilitate immediate use by the research community. The significance of PandaSet lies in its potential to improve the robustness of autonomous driving perception systems by providing richer, more diverse, and higher-precision data. By offering a free commercial license and a unique combination of mechanical spinning and forward-facing LiDARs, it lowers barriers to entry for both academic and industrial research. The extensive annotation taxonomy and coverage of challenging driving scenarios address previous limitations in dataset diversity, enabling the development of algorithms capable of handling complex, real-world environments. This resource is expected to accelerate advancements in machine learning models for autonomous vehicles by providing a more representative ground truth for training and evaluation.
Provenance
The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.
| Stage | Outcome | Tool | Model | Prompt | Attempts | Completed |
|---|---|---|---|---|---|---|
| discover | success | OpenAlex-citations | — | — | 1 | 2026-06-20 |
| archive | success | semantic_scholar | — | — | 6 | 2026-06-26 |
| extract | success | cached | — | — | 2 | 2026-06-26 |
| clean | success | clean | — | — | 1 | 2026-06-20 |
| chunk | success | chunk | — | — | 1 | 2026-06-20 |
| embed | success | embed | Qwen/Qwen3-Embedding-8B | — | 1 | 2026-06-20 |
| promote | success | — | — | — | 1 | 2026-06-20 |
| summarize | success | llm | qwen3.6-27b-prismaquant | summ-v5 | 1 | 2026-06-26 |
| tag | success | vector_similarity | — | — | 6 | 2026-06-20 |
| verify | success | — | — | — | 1 | 2026-06-26 |
Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.
Topics
Ranked by relevance to this paper. Hover a topic for its definition.
Information type
What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).
- Methodological Resource: dataset resource, tool software