We propose a deep convolutional neural network architecture codenamed Inception that achieves the new state of the art for classification and detection in the ImageNet Large-Scale Visual Recognition Challenge 2014 (ILSVRC14). The main hallmark of this architecture is the improved utilization of the computing resources inside the network. By a carefully crafted design, we increased the depth and width of the network while keeping the computational budget constant. To optimize quality, the architectural decisions were based on the Hebbian principle and the intuition of multi-scale processing. One particular incarnation used in our submission for ILSVRC14 is called GoogLeNet, a 22 layers deep network, the quality of which is assessed in the context of classification and detection.

/pdf/going-deeper-with-convolutions-1yobw2o2ds.pdf

Going deeper with convolutions

We present ORB-SLAM2, a complete simultaneous localization and mapping (SLAM) system for monocular, stereo and RGB-D cameras, including map reuse, loop closing, and relocalization capabilities. The system works in real time on standard central processing units in a wide variety of environments from small hand-held indoors sequences, to drones flying in industrial environments and cars driving around a city. Our back-end, based on bundle adjustment with monocular and stereo observations, allows for accurate trajectory estimation with metric scale. Our system includes a lightweight localization mode that leverages visual odometry tracks for unmapped regions and matches with map points that allow for zero-drift localization. The evaluation on 29 popular public sequences shows that our method achieves state-of-the-art accuracy, being in most cases the most accurate SLAM solution. We publish the source code, not only for the benefit of the SLAM community, but with the aim of being an out-of-the-box SLAM solution for researchers in other fields.

ORB-SLAM2: An Open-Source SLAM System for Monocular, Stereo, and RGB-D Cameras

We present ORB-SLAM2 a complete SLAM system for monocular, stereo and RGB-D cameras, including map reuse, loop closing and relocalization capabilities. The system works in real-time on standard CPUs in a wide variety of environments from small hand-held indoors sequences, to drones flying in industrial environments and cars driving around a city. Our back-end based on bundle adjustment with monocular and stereo observations allows for accurate trajectory estimation with metric scale. Our system includes a lightweight localization mode that leverages visual odometry tracks for unmapped regions and matches to map points that allow for zero-drift localization. The evaluation on 29 popular public sequences shows that our method achieves state-of-the-art accuracy, being in most cases the most accurate SLAM solution. We publish the source code, not only for the benefit of the SLAM community, but with the aim of being an out-of-the-box SLAM solution for researchers in other fields.

ORB-SLAM2: an Open-Source SLAM System for Monocular, Stereo and RGB-D Cameras

A key requirement for leveraging supervised deep learning methods is the availability of large, labeled datasets. Unfortunately, in the context of RGB-D scene understanding, very little data is available &#x2013; current datasets cover a small range of scene views and have limited semantic annotations. To address this issue, we introduce ScanNet, an RGB-D video dataset containing 2.5M views in 1513 scenes annotated with 3D camera poses, surface reconstructions, and semantic segmentations. To collect this data, we designed an easy-to-use and scalable RGB-D capture system that includes automated surface reconstruction and crowdsourced semantic annotation. We show that using this data helps achieve state-of-the-art performance on several 3D scene understanding tasks, including 3D object classification, semantic voxel labeling, and CAD model retrieval.

/pdf/scannet-richly-annotated-3d-reconstructions-of-indoor-scenes-37ihdazbor.pdf

ScanNet: Richly-Annotated 3D Reconstructions of Indoor Scenes

Three-dimensional models provide a volumetric representation of space which is important for a variety of robotic applications including flying robots and robots that are equipped with manipulators. In this paper, we present an open-source framework to generate volumetric 3D environment models. Our mapping approach is based on octrees and uses probabilistic occupancy estimation. It explicitly represents not only occupied space, but also free and unknown areas. Furthermore, we propose an octree map compression method that keeps the 3D models compact. Our framework is available as an open-source C++ library and has already been successfully applied in several robotics projects. We present a series of experimental results carried out with real robots and on publicly available real-world datasets. The results demonstrate that our approach is able to update the representation efficiently and models the data consistently while keeping the memory requirement at a minimum.

/pdf/octomap-an-efficient-probabilistic-3d-mapping-framework-3dp42nnofi.pdf

OctoMap: an efficient probabilistic 3D mapping framework based on octrees

We introduce the Imperial College London and
National University of Ireland Maynooth (ICL-NUIM) dataset
for the evaluation of visual odometry, 3D reconstruction and
SLAM algorithms that typically use RGB-D data. We present
a collection of handheld RGB-D camera sequences within
synthetically generated environments. RGB-D sequences with
perfect ground truth poses are provided as well as a ground
truth surface model that enables a method of quantitatively
evaluating the final map or surface reconstruction accuracy.
Care has been taken to simulate typically observed real-world
artefacts in the synthetic imagery by modelling sensor noise in
both RGB and depth data. While this dataset is useful for the
evaluation of visual odometry and SLAM trajectory estimation,
our main focus is on providing a method to benchmark the
surface reconstruction accuracy which to date has been missing
in the RGB-D community despite the plethora of ground truth
RGB-D datasets available.

/pdf/a-benchmark-for-rgb-d-visual-odometry-3d-reconstruction-and-2zy6prul21.pdf

A benchmark for RGB-D visual odometry, 3D reconstruction and SLAM

We present a novel approach to real-time dense visual SLAM. Our system is capable of capturing comprehensive dense globally consistent surfel-based maps of room scale environments explored using an RGB-D camera in an incremental online fashion, without pose graph optimisation or any postprocessing steps. This is accomplished by using dense frame-tomodel camera tracking and windowed surfel-based fusion coupled with frequent model refinement through non-rigid surface deformations. Our approach applies local model-to-model surface loop closure optimisations as often as possible to stay close to the mode of the map distribution, while utilising global loop closure to recover from arbitrary drift and maintain global consistency.

http://www.roboticsproceedings.org/rss11/p01.pdf

ElasticFusion: Dense SLAM Without A Pose Graph

In this paper we present an extension to the
KinectFusion algorithm that permits dense mesh-based mapping of extended scale environments in real-time. This is achieved through (i) altering the original algorithm such that the region of space being mapped by the KinectFusion algorithm can vary dynamically, (ii) extracting a dense point cloud from the regions that leave the KinectFusion volume due to this variation, and, (iii) incrementally adding the resulting points to a triangular mesh representation of the environment. The system is implemented as a set of hierarchical multi-threaded components which are capable of operating in real-time. The architecture facilitates the creation and integration of new modules with minimal impact on the performance on the dense volume tracking and surface reconstruction modules. We provide experimental results demonstrating the system’s ability to map areas considerably
beyond the scale of the original KinectFusion algorithm including a two story apartment and an extended sequence taken from a car at night. In order to overcome failure of the iterative closest point (ICP) based odometry in areas of low geometric features we have evaluated the Fast Odometry from Vision (FOVIS) system as an alternative. We provide a comparison between the two approaches where we show a trade off between the reduced drift of the visual odometry approach and the higher local mesh quality of the ICP-based approach. Finally we present ongoing work on incorporating full simultaneous localisation and mapping (SLAM) pose-graph optimisation.

/pdf/kintinuous-spatially-extended-kinectfusion-44w0up5ji0.pdf

Kintinuous: Spatially Extended KinectFusion

We present a novel approach to real-time dense visual simultaneous localisation and mapping. Our system is capable of capturing comprehensive dense globally consistent surfel-based maps of room scale environments and beyond explored using an RGB-D camera in an incremental online fashion, without pose graph optimization or any post-processing steps. This is accomplished by using dense frame-to-model camera tracking and windowed surfel-based fusion coupled with frequent model refinement through non-rigid surface deformations. Our approach applies local model-to-model surface loop closure optimizations as often as possible to stay close to the mode of the map distribution, while utilizing global loop closure to recover from arbitrary drift and maintain global consistency. In the spirit of improving map quality as well as tracking accuracy and robustness, we furthermore explore a novel approach to real-time discrete light source detection. This technique is capable of detecting numerous light sources in indoo...

ElasticFusion: Real-time dense SLAM and light source estimation

We present a new simultaneous localization and mapping SLAM system capable of producing high-quality globally consistent surface reconstructions over hundreds of meters in real time with only a low-cost commodity RGB-D sensor. By using a fused volumetric surface reconstruction we achieve a much higher quality map over what would be achieved using raw RGB-D point clouds. In this paper we highlight three key techniques associated with applying a volumetric fusion-based mapping system to the SLAM problem in real time. First, the use of a GPU-based 3D cyclical buffer trick to efficiently extend dense every-frame volumetric fusion of depth maps to function over an unbounded spatial region. Second, overcoming camera pose estimation limitations in a wide variety of environments by combining both dense geometric and photometric camera pose constraints. Third, efficiently updating the dense map according to place recognition and subsequent loop closure constraints by the use of an 'as-rigid-as-possible' space deformation. We present results on a wide variety of aspects of the system and show through evaluation on de facto standard RGB-D benchmarks that our system performs strongly in terms of trajectory estimation, map quality and computational performance in comparison to other state-of-the-art systems.

/pdf/real-time-large-scale-dense-rgb-d-slam-with-volumetric-nsvs7jcopb.pdf

Thomas Whelan

Papers

A benchmark for RGB-D visual odometry, 3D reconstruction and SLAM

ElasticFusion: Dense SLAM Without A Pose Graph

Kintinuous: Spatially Extended KinectFusion

ElasticFusion: Real-time dense SLAM and light source estimation

Real-time large-scale dense RGB-D SLAM with volumetric fusion