Point Cloud Registration

KITTI (Karlsruhe Institute of Technology and Toyota Technological Institute) is one of the most popular datasets for use in mobile robotics and autonomous driving. It consists of hours of traffic scenarios recorded with a variety of sensor modalities, including high-resolution RGB, grayscale stereo cameras, and a 3D laser scanner. Despite its popularity, the dataset itself does not contain ground truth for semantic segmentation. However, various researchers have manually annotated parts of the dataset to fit their necessities. Álvarez et al. generated ground truth for 323 images from the road detection challenge with three classes: road, vertical, and sky. Zhang et al. annotated 252 (140 for training and 112 for testing) acquisitions – RGB and Velodyne scans – from the tracking challenge for ten object categories: building, sky, road, vegetation, sidewalk, car, pedestrian, cyclist, sign/pole, and fence. Ros et al. labeled 170 training images and 46 testing images (from the visual odometry challenge) with 11 classes: building, tree, sky, car, sign, road, pedestrian, fence, pole, sidewalk, and bicyclist. Source: A Review on Deep Learning Techniques Applied to Semantic Segmentation Image Source: http://www.cvlibs.net/datasets/kitti/eval_object.php?obj_benchmark=3d

The 3DMATCH benchmark evaluates how well descriptors (both 2D and 3D) can establish correspondences between RGB-D frames of different views. The dataset contains 2D RGB-D patches and 3D patches (local TDF voxel grid volumes) of wide-baselined correspondences. The pixel size of each 2D patch is determined by the projection of the 0.3m3 local 3D patch around the interest point onto the image plane. Source: 3DMatch: Learning Local Geometric Descriptors from RGB-D Reconstructions

A novel dataset and benchmark, which features 1482 RGB-D scans of 478 environments across multiple time steps. Each scene includes several objects whose positions change over time, together with ground truth annotations of object instances and their respective 6DoF mappings among re-scans. Source: RIO: 3D Object Instance Re-Localization in Changing Indoor Environments

DeformingThings4D is a synthetic dataset containing 1,972 animation sequences spanning 31 categories of humanoids and animals. It provides 200 animations for humanoids and 1772 animations for animals. #### Use case of the dataset The dataset is designed to tackle the following tasks using data-driven approaches: - Scene flow estimation - Non-rigid tracking/registration - Shape and motion completion - Learning riggings from observation - Generic non-rigid reconstruction Source: DeformingThings4D dataset Image source: DeformingThings4D dataset

ScanNet++ is a large scale dataset with 450+ 3D indoor scenes containing sub-millimeter resolution laser scans, registered 33-megapixel DSLR images, and commodity RGB-D streams from iPhone. The 3D reconstructions are annotated with long-tail and label-ambiguous semantics to benchmark semantic understanding methods, while the coupled DSLR and iPhone captures enable benchmarking of novel view synthesis methods in high-quality and commodity settings.

UrbanLoco is a mapping/localization dataset collected in highly-urbanized environments with a full sensor-suite. The dataset includes 13 trajectories collected in San Francisco and Hong Kong, covering a total length of over 40 kilometers. Source: UrbanLoco: A Full Sensor Suite Dataset for Mapping and Localization in Urban Scenes Image Source: https://advdataset2019.wixsite.com/urbanloco

A benchmark for matching and registration of partial point clouds with time-varying geometry. It is constructed using randomly selected 1761 sequences from DeformingThings4D.

Cross-source point cloud dataset for registration task. It includes point clouds from structure from motion (SFM), Kinect, Lidar.

FPv1 (prior name FAUST-partial) is a 3D registration benchmark dataset created to address the lack of data variability in the existing 3D registration benchmarks such as: 3DMatch, ETH, KITTI. The original FAUST training dataset is comprised of 100 3D scans of human bodies. The benchmark generation for a single scan from the FAUST training dataset can be summarized as follows: 1. Make xz-plane the floor by translating the minimal bounding box point of the scan to the origin 2. Surround the scan with a regular icosahaedron. Each point of the icosahaedron acts as a viewpoint 3. For each viewpoint, create a partial point cloud using the hidden point removal algorithm Finally, for a pair of partial point clouds with the desired overalp, generate a random rotation from the desired rotation range and translation range.

FAUST-partial is a 3D registration benchmark dataset created to provide a more informative evaluation of 3D registration methods. The dataset addresses two main limitations of current 3D registration benchmarks: 1. Lack of data variability within the current registration benchmarks 2. Capability to evaluate the registration method on a single 3D registration parameter (rotation, translation or overlap) The benchmark is created using the FAUST training dataset comprised of 100 3D scans of human bodies. Note however, that the methodology can be applied to any point cloud dataset. The benchmark generation for a single scan from the FAUST training dataset can be summarized as follows: - Surround the scan with a regular icosahaedron. Each point of the icosahaedron acts as a viewpoint - For each viewpoint, create a partial point cloud using the hidden point removal algorithm - Finally, for a pair of partial point clouds in a desired range of overlap, generate a random rotation from the desired rotation and translation ranges. To create a more informative 3D registration benchmark, 3 difficulty settings are used: Easy (E), Medium (M) and Hard (H) for each registration parameter: Rotation (R), Translation (T) and Overlap (O). This way, 9 benchmarks are created in which each triplet increases the difficulty of one registration parameter whilst fixing the difficulty of the other two. This allows to evaluate a 3D registration method w.r.t. only one registration parameter and observe the robustness of the method w.r.t. that parameter. The benchmarks are denominated as FP-{R,T,O}-{E,M,H}.

This dataset presents a vision and perception research dataset collected in Rome, featuring RGB data, 3D point clouds, IMU, and GPS data. We introduce a new benchmark targeting visual odometry and SLAM, to advance the research in autonomous robotics and computer vision. This work complements existing datasets by simultaneously addressing several issues, such as environment diversity, motion patterns, and sensor frequency. It uses up-to-date devices and presents effective procedures to accurately calibrate the intrinsic and extrinsic of the sensors while addressing temporal synchronization. During recording, we cover multi-floor buildings, gardens, urban and highway scenarios. Combining handheld and car-based data collections, our setup can simulate any robot (quadrupeds, quadrotors, autonomous vehicles). The dataset includes an accurate 6-dof ground truth based on a novel methodology that refines the RTK-GPS estimate with LiDAR point clouds through Bundle Adjustment. All sequences divided in training and testing are accessible through our website.

The Cooperative Driving dataset is a synthetic dataset generated using CARLA that contains lidar data from multiple vehicles navigating simultaneously through a diverse set of driving scenarios. This dataset was created to enable further research in multi-agent perception (cooperative perception) including cooperative 3D object detection, cooperative object tracking, multi-agent SLAM and point cloud registration. Towards that goal, all the frames have been labelled with ground-truth sensor pose and 3D object bounding boxes.