Volumetric Instance-Aware Semantic Mapping and 3D Object Discovery

Year: 2019
DOI: 10.3929/ethz-c-000788341

RGB-D rosbags for the Voxblox++ (Volumetric Instance-Aware Semantic Mapping) paper, including a SceneNN sequence and an office-floor run with multiple depth/RGB streams for online mapping experiments.

Dataset Details

• SceneNN 231: 3:06 duration, 7.9 GB; topics include RGB-D images, camera info, TF
• Office floor (ASL): 5:12 duration, 6.3 GB; two Primesense cameras (front/table) with depth_registered and RGB streams; high-frequency TF
• Suitable for semantic/instance-aware volumetric mapping and object discovery benchmarks

Downloads

• SceneNN sequence: https://doi.org/10.3929/ethz-c-000788341 (was scenenn_231.bag, 7.9 GB)
• Office floor sequence: https://doi.org/10.3929/ethz-c-000788341 (was asl_office_floor.bag, 6.3 GB)

Citation

M. Grinvald, F. Furrer, T. Novkovic, J. J. Chung, C. Cadena, R. Siegwart, J. Nieto,
“Volumetric Instance-Aware Semantic Mapping and 3D Object Discovery”, IEEE Robotics and Automation Letters (RA-L), 2019.

Tags: rgbd, mapping, semantic, instance, indoor