Imperial College London

ProfessorAndrewDavison

Faculty of EngineeringDepartment of Computing

Professor of Robot Vision
 
 
 
//

Contact

 

+44 (0)20 7594 8316a.davison Website

 
 
//

Assistant

 

Ms Lucy Atthis +44 (0)20 7594 8259

 
//

Location

 

303William Penney LaboratorySouth Kensington Campus

//

Summary

 

Publications

Publication Type
Year
to

152 results found

Ortiz J, Evans T, Sucar E, Davison AJet al., 2022, Incremental abstraction in distributed probabilistic SLAM graphs, 2022 IEEE International Conference on Robotics and Automation (ICRA), Publisher: IEEE

Scene graphs represent the key components of a scene in a compact and semantically rich way, but are difficult to build during incremental SLAM operation because of the challenges of robustly identifying abstract scene elements and optimising continually changing, complex graphs. We present a distributed, graph-based SLAM framework for incrementally building scene graphs based on two novel components. First, we propose an incremental abstraction framework in which a neural network proposes abstract scene elements that are incorporated into the factor graph of a feature-based monocular SLAM system. Scene elements are confirmed or rejected through optimisation and incrementally replace the points yielding a more dense, semantic and compact representation. Second, enabled by our novel routing procedure, we use Gaussian Belief Propagation (GBP) for distributed inference on a graph processor. The time per iteration of GBP is structure-agnostic and we demonstrate the speed advantages over direct methods for inference of heterogeneous factor graphs. We run our system on real indoor datasets using planar abstractions and recover the major planes with significant compression.

Conference paper

Wada K, James S, Davison AJ, 2022, SafePicking: learning safe object extraction via object-level mapping, 2022 IEEE International Conference on Robotics and Automation (ICRA), Publisher: IEEE

Robots need object-level scene understanding to manipulate objects while reasoning about contact, support, and occlusion among objects. Given a pile of objects, object recognition and reconstruction can identify the boundary of object instances, giving important cues as to how the objects form and support the pile. In this work, we present a system, SafePicking, that integrates object-level mapping and learning-based motion planning to generate a motion that safely extracts occluded target objects from a pile. Planning is done by learning a deep Q-network that receives observations of predicted poses and a depth-based heightmap to output a motion trajectory, trained to maximize a safety metric reward. Our results show that the observation fusion of poses and depth-sensing gives both better performance and robustness to the model. We evaluate our methods using the YCB objects in both simulation and the real world, achieving safe object extraction from piles.

Conference paper

Wada K, James S, Davison AJ, 2022, ReorientBot: learning object reorientation for specific-posed placement, 2022 IEEE International Conference on Robotics and Automation (ICRA), Publisher: IEEE

Robots need the capability of placing objects in arbitrary, specific poses to rearrange the world and achieve various valuable tasks. Object reorientation plays a crucial role in this as objects may not initially be oriented such that the robot can grasp and then immediately place them in a specific goal pose. In this work, we present a vision-based manipulation system, ReorientBot, which consists of 1) visual scene understanding with pose estimation and volumetric reconstruction using an onboard RGB-D camera; 2) learned waypoint selection for successful and efficient motion generation for reorientation; 3) traditional motion planning to generate a collision-free trajectory from the selected waypoints. We evaluate our method using the YCB objects in both simulation and the real world, achieving 93% overall success, 81% improvement in success rate, and 22% improvement in execution time compared to a heuristic approach. We demonstrate extended multi-object rearrangement showing the general capability of the system.

Conference paper

James S, Davison AJ, 2022, Q-Attention: Enabling Efficient Learning for Vision-Based Robotic Manipulation, IEEE ROBOTICS AND AUTOMATION LETTERS, Vol: 7, Pages: 1612-1619, ISSN: 2377-3766

Journal article

Scona R, Matsuki H, Davison A, 2022, From scene flow to visual odometry through local and global regularisation in markov random fields, IEEE Robotics and Automation Letters, Vol: 7, Pages: 4299-4306, ISSN: 2377-3766

We revisit pairwise Markov Random Field (MRF) formulations for RGB-D scene flow and leverage novel advances in processor design for real-time implementations. We consider scene flow approaches which consist of data terms enforcing intensity consistency between consecutive images, together with regularisation terms which impose smoothness over the flow field. To achieve real-time operation, previous systems leveraged GPUs and implemented regularisation only between variables corresponding to neighbouring pixels. Such systems could estimate continuously deforming flow fields but the lack of global regularisation over the whole field made them ineffective for visual odometry. We leverage the GraphCore Intelligence Processing Unit (IPU) graph processor chip, which consists of 1216 independent cores called tiles, each with 256 kB local memory. The tiles are connected to an ultrafast all-to-all communication fabric which enables efficient data transmission between the tiles in an arbitrary communication pattern. We propose a distributed formulation for dense RGB-D scene flow based on Gaussian Belief Propagation which leverages the architecture of this processor to implement both local and global regularisation. Local regularisation is enforced for pairs of flow estimates whose corresponding pixels are neighbours, while global regularisation is defined for flow estimate pairs whose corresponding pixels are far from each other on the image plane. Using both types of regularisation allows our algorithm to handle a variety of in-scene motion and makes it suitable for estimating deforming scene flow, piece-wise rigid scene flow and visual odometry within the same system.

Journal article

Sucar E, Liu S, Ortiz J, Davison AJet al., 2022, iMAP: implicit mapping and positioning in real-time, 2021 IEEE/CVF International Conference on Computer Vision (ICCV), Publisher: IEEE, Pages: 6209-6218

We show for the first time that a multilayer perceptron (MLP) can serve as the only scene representation in a real-time SLAM system for a handheld RGB-D camera. Our network is trained in live operation without prior data, building a dense, scene-specific implicit 3D model of occupancy and colour which is also immediately used for tracking.Achieving real-time SLAM via continual training of a neural network against a live image stream requires significant innovation. Our iMAP algorithm uses a keyframe structure and multi-processing computation flow, with dynamic information-guided pixel sampling for speed, with tracking at 10 Hz and global map updating at 2 Hz. The advantages of an implicit MLP over standard dense SLAM techniques include efficient geometry representation with automatic detail control and smooth, plausible filling-in of unobserved regions such as the back surfaces of objects.

Conference paper

Zhi S, Laidlow T, Leutenegger S, Davison AJet al., 2022, In-place scene labelling and understanding with implicit scene representation, 2021 IEEE/CVF International Conference on Computer Vision (ICCV), Publisher: IEEE

Semantic labelling is highly correlated with geometry and radiance reconstruction, as scene entities with similar shape and appearance are more likely to come from similar classes. Recent implicit neural reconstruction techniques are appealing as they do not require prior training data, but the same fully self-supervised approach is not possible for semantics because labels are human-defined properties.We extend neural radiance fields (NeRF) to jointly encode semantics with appearance and geometry, so that complete and accurate 2D semantic labels can be achieved using a small amount of in-place annotations specific to the scene. The intrinsic multi-view consistency and smoothness of NeRF benefit semantics by enabling sparse labels to efficiently propagate. We show the benefit of this approach when labels are either sparse or very noisy in room-scale scenes. We demonstrate its advantageous properties in various interesting applications such as an efficient scene labelling tool, novel semantic view synthesis, label denoising, super-resolution, label interpolation and multi-view semantic label fusion in visual semantic mapping systems.

Conference paper

Landgraf Z, Scona R, Laidlow T, James S, Leutenegger S, Davison AJet al., 2022, SIMstack: a generative shape and instance model for unordered object stacks, 2021 IEEE/CVF International Conference on Computer Vision (ICCV), Publisher: IEEE

By estimating 3D shape and instances from a single view, we can capture information about an environment quickly, without the need for comprehensive scanning and multi-view fusion. Solving this task for composite scenes (such as object stacks) is challenging: occluded areas are not only ambiguous in shape but also in instance segmentation; multiple decompositions could be valid. We observe that physics constrains decomposition as well as shape in occluded regions and hypothesise that a latent space learned from scenes built under physics simulation can serve as a prior to better predict shape and instances in occluded regions. To this end we propose SIMstack, a depth-conditioned Variational Auto-Encoder (VAE), trained on a dataset of objects stacked under physics simulation. We formulate instance segmentation as a centre voting task which allows for class-agnostic detection and doesn’t require setting the maximum number of objects in the scene. At test time, our model can generate 3D shape and instance segmentation from a single depth view, probabilistically sampling proposals for the occluded region from the learned latent space. Our method has practical applications in providing robots some of the ability humans have to make rapid intuitive inferences of partially observed scenes. We demonstrate an application for precise (non-disruptive) object grasping of unknown objects from a single depth view.

Conference paper

James S, Wada K, Laidlow T, Davison AJet al., 2022, Coarse-to-Fine Q-attention: Efficient Learning for Visual Robotic Manipulation via Discretisation, Pages: 13729-13738, ISSN: 1063-6919

We present a coarse-to-fine discretisation method that enables the use of discrete reinforcement learning approaches in place of unstable and data-inefficient actorcritic methods in continuous robotics domains. This approach builds on the recently released ARM algorithm, which replaces the continuous next-best pose agent with a discrete one, with coarse-to-fine Q-attention. Given a voxelised scene, coarse-to-fine Q-attention learns what part of the scene to 'zoom' into. When this 'zooming' behaviour is applied iteratively, it results in a near-lossless discretisation of the translation space, and allows the use of a discrete action, deep Q-learning method. We show that our new coarse-to-fine algorithm achieves state-of-the-art performance on several difficult sparsely rewarded RLBench vision-based robotics tasks, and can train real-world policies, tabula rasa, in a matter of minutes, with as little as 3 demonstrations.

Conference paper

Gallego G, Delbruck T, Orchard GM, Bartolozzi C, Taba B, Censi A, Leutenegger S, Davison A, Conradt J, Daniilidis K, Scaramuzza Det al., 2022, Event-Based Vision: A Survey, Publisher: IEEE COMPUTER SOC

Working paper

Matsuki H, Scona R, Czarnowski J, Davison AJet al., 2021, CodeMapping: real-time dense mapping for sparse SLAM using compact scene representations, IEEE Robotics and Automation Letters, Vol: 6, Pages: 7105-7112, ISSN: 2377-3766

We propose a novel dense mapping framework for sparse visual SLAM systems which leverages a compact scene representation. State-of-the-art sparse visual SLAM systems provide accurate and reliable estimates of the camera trajectory and locations of landmarks. While these sparse maps are useful for localization, they cannot be used for other tasks such as obstacle avoidance or scene understanding. In this letter we propose a dense mapping framework to complement sparse visual SLAM systems which takes as input the camera poses, keyframes and sparse points produced by the SLAM system and predicts a dense depth image for every keyframe. We build on CodeSLAM [1] and use a variational autoencoder (VAE) which is conditioned on intensity, sparse depth and reprojection error images from sparse SLAM to predict an uncertainty-aware dense depth map. The use of a VAE then enables us to refine the dense depth images through multi-view optimization which improves the consistency of overlapping frames. Our mapper runs in a separate thread in parallel to the SLAM system in a loosely coupled manner. This flexible design allows for integration with arbitrary metric sparse SLAM systems without delaying the main SLAM process. Our dense mapper can be used not only for local mapping but also globally consistent dense 3D reconstruction through TSDF fusion. We demonstrate our system running with ORB-SLAM3 and show accurate dense depth estimation which could enable applications such as robotics and augmented reality.

Journal article

Xu B, Davison AJ, Leutenegger S, 2020, Deep probabilistic feature-metric tracking, Publisher: arXiv

Dense image alignment from RGB-D images remains a critical issue forreal-world applications, especially under challenging lighting conditions andin a wide baseline setting. In this paper, we propose a new framework to learna pixel-wise deep feature map and a deep feature-metric uncertainty mappredicted by a Convolutional Neural Network (CNN), which together formulate adeep probabilistic feature-metric residual of the two-view constraint that canbe minimised using Gauss-Newton in a coarse-to-fine optimisation framework.Furthermore, our network predicts a deep initial pose for faster and morereliable convergence. The optimisation steps are differentiable and unrolled totrain in an end-to-end fashion. Due to its probabilistic essence, our approachcan easily couple with other residuals, where we show a combination with ICP.Experimental results demonstrate state-of-the-art performance on the TUM RGB-Ddataset and 3D rigid object tracking dataset. We further demonstrate ourmethod's robustness and convergence qualitatively.

Working paper

Wada K, Sucar E, James S, Lenton D, Davison AJet al., 2020, MoreFusion: multi-object reasoning for 6D pose estimation from volumetric fusion, 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Publisher: IEEE

Robots and other smart devices need efficient object-based scene representations from their on-board vision systems to reason about contact, physics and occlusion. Recognized precise object models will play an important role alongside non-parametric reconstructions of unrecognized structures. We present a system which can estimate the accurate poses of multiple known objects in contact and occlusion from real-time, embodied multi-view vision. Our approach makes 3D object pose proposals from single RGB-D views, accumulates pose estimates and non-parametric occupancy information from multiple views as the camera moves, and performs joint optimization to estimate consistent, non-intersecting poses for multiple objects in contact. We verify the accuracy and robustness of our approach experimentally on 2 object datasets: YCB-Video, and our own challenging Cluttered YCB-Video. We demonstrate a real-time robotics application where a robot arm precisely and orderly disassembles complicated piles of objects, using only on-board RGB-D vision.

Conference paper

James S, Ma Z, Arrojo DR, Davison AJet al., 2020, RLBench: The robot learning benchmark & learning environment, IEEE Robotics and Automation Letters, Vol: 5, Pages: 3019-3026, ISSN: 2377-3766

We present a challenging new benchmark and learning-environment for robot learning: RLBench. The benchmark features 100 completely unique, hand-designed tasks, ranging in difficulty from simple target reaching and door opening to longer multi-stage tasks, such as opening an oven and placing a tray in it. We provide an array of both proprioceptive observations and visual observations, which include rgb, depth, and segmentation masks from an over-the-shoulder stereo camera and an eye-in-hand monocular camera. Uniquely, each task comes with an infinite supply of demos through the use of motion planners operating on a series of waypoints given during task creation time; enabling an exciting flurry of demonstration-based learning possibilities. RLBench has been designed with scalability in mind; new tasks, along with their motion-planned demos, can be easily created and then verified by a series of tools, allowing users to submit their own tasks to the RLBench task repository. This large-scale benchmark aims to accelerate progress in a number of vision-guided manipulation research areas, including: reinforcement learning, imitation learning, multi-task learning, geometric computer vision, and in particular, few-shot learning. With the benchmark's breadth of tasks and demonstrations, we propose the first large-scale few-shot challenge in robotics. We hope that the scale and diversity of RLBench offers unparalleled research opportunities in the robot learning community and beyond. Benchmarking code and videos can be found at https://sites.google.com/view/rlbench .

Journal article

Bonardi A, James S, Davison AJ, 2020, Learning one-shot imitation from humans without humans, IEEE Robotics and Automation Letters, Vol: 5, Pages: 3533-3539, ISSN: 2377-3766

Humans can naturally learn to execute a new task by seeing it performed by other individuals once, and then reproduce it in a variety of configurations. Endowing robots with this ability of imitating humans from third person is a very immediate and natural way of teaching new tasks. Only recently, through meta-learning, there have been successful attempts to one-shot imitation learning from humans; however, these approaches require a lot of human resources to collect the data in the real world to train the robot. But is there a way to remove the need for real world human demonstrations during training? We show that with Task-Embedded Control Networks, we can infer control polices by embedding human demonstrations that can condition a control policy and achieve one-shot imitation learning. Importantly, we do not use a real human arm to supply demonstrations during training, but instead leverage domain randomisation in an application that has not been seen before: sim-to-real transfer on humans. Upon evaluating our approach on pushing and placing tasks in both simulation and in the real world, we show that in comparison to a system that was trained on real-world data we are able to achieve similar results by utilising only simulation data. Videos can be found here: https://sites.google.com/view/tecnets-humans .

Journal article

Ortiz J, Pupilli M, Leutenegger S, Davison AJet al., 2020, Bundle adjustment on a graph processor, Publisher: arXiv

Graph processors such as Graphcore's Intelligence Processing Unit (IPU) arepart of the major new wave of novel computer architecture for AI, and have ageneral design with massively parallel computation, distributed on-chip memoryand very high inter-core communication bandwidth which allows breakthroughperformance for message passing algorithms on arbitrary graphs. We show for thefirst time that the classical computer vision problem of bundle adjustment (BA)can be solved extremely fast on a graph processor using Gaussian BeliefPropagation. Our simple but fully parallel implementation uses the 1216 coreson a single IPU chip to, for instance, solve a real BA problem with 125keyframes and 1919 points in under 40ms, compared to 1450ms for the Ceres CPUlibrary. Further code optimisation will surely increase this difference onstatic problems, but we argue that the real promise of graph processing is forflexible in-place optimisation of general, dynamically changing factor graphsrepresenting Spatial AI problems. We give indications of this with experimentsshowing the ability of GBP to efficiently solve incremental SLAM problems, anddeal with robust cost functions and different types of factors.

Working paper

Bloesch M, Laidlow T, Clark R, Leutenegger S, Davison Aet al., 2020, Learning meshes for dense visual SLAM, 2019 IEEE/CVF International Conference on Computer Vision (ICCV), Publisher: IEEE

Estimating motion and surrounding geometry of a moving camera remains a challenging inference problem. From an information theoretic point of view, estimates should get better as more information is included, such as is done in dense SLAM, but this is strongly dependent on the validity of the underlying models. In the present paper, we use triangular meshes as both compact and dense geometry representation. To allow for simple and fast usage, we propose a view-based formulation for which we predict the in-plane vertex coordinates directly from images and then employ the remaining vertex depth components as free variables. Flexible and continuous integration of information is achieved through the use of a residual based inference technique. This so-called factor graph encodes all information as mapping from free variables to residuals, the squared sum of which is minimised during inference. We propose the use of different types of learnable residuals, which are trained end-to-end to increase their suitability as information bearing models and to enable accurate and reliable estimation. Detailed evaluation of all components is provided on both synthetic and real data which confirms the practicability of the presented approach.

Conference paper

Landgraf Z, Falck F, Bloesch M, Leutenegger S, Davison Aet al., 2020, Comparing view-based and map-based semantic labelling in real-time SLAM, Publisher: arXiv

Generally capable Spatial AI systems must build persistent scenerepresentations where geometric models are combined with meaningful semanticlabels. The many approaches to labelling scenes can be divided into two cleargroups: view-based which estimate labels from the input view-wise data and thenincrementally fuse them into the scene model as it is built; and map-basedwhich label the generated scene model. However, there has so far been noattempt to quantitatively compare view-based and map-based labelling. Here, wepresent an experimental framework and comparison which uses real-time heightmap fusion as an accessible platform for a fair comparison, opening up theroute to further systematic research in this area.

Working paper

Czarnowski J, Laidlow T, Clark R, Davison AJet al., 2020, DeepFactors: Real-time probabilistic dense monocular SLAM, IEEE Robotics and Automation Letters, Vol: 5, Pages: 721-728, ISSN: 2377-3766

The ability to estimate rich geometry and camera motion from monocular imagery is fundamental to future interactive robotics and augmented reality applications. Different approaches have been proposed that vary in scene geometry representation (sparse landmarks, dense maps), the consistency metric used for optimising the multi-view problem, and the use of learned priors. We present a SLAM system that unifies these methods in a probabilistic framework while still maintaining real-time performance. This is achieved through the use of a learned compact depth map representation and reformulating three different types of errors: photometric, reprojection and geometric, which we make use of within standard factor graph software. We evaluate our system on trajectory estimation and depth reconstruction on real-world sequences and present various examples of estimated dense geometry.

Journal article

Johns E, Liu S, Davison A, 2020, End-to-end multi-task learning with attention, The IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2019, Publisher: IEEE

We propose a novel multi-task learning architecture, which allows learning of task-specific feature-level attention. Our design, the Multi-Task Attention Network (MTAN), consists of a single shared network containing a global feature pool, together with a soft-attention module for each task. These modules allow for learning of task-specific features from the global features, whilst simultaneously allowing for features to be shared across different tasks. The architecture can be trained end-to-end and can be built upon any feed-forward neural network, is simple to implement, and is parameter efficient. We evaluate our approach on a variety of datasets, across both image-to-image predictions and image classification tasks. We show that our architecture is state-of-the-art in multi-task learning compared to existing methods, and is also less sensitive to various weighting schemes in the multi-task loss function. Code is available at https://github.com/lorenmt/mtan.

Conference paper

Sucar E, Wada K, Davison A, 2020, NodeSLAM: Neural Object Descriptors for Multi-View Shape Reconstruction, 8th International Conference on 3D Vision (3DV), Publisher: IEEE, Pages: 949-958, ISSN: 2378-3826

Conference paper

Liu S, Davison A, Johns E, 2019, Self-supervised generalisation with meta auxiliary learning, 33rd Conference on Neural Information Processing Systems (NeurIPS 2019), Publisher: Neural Information Processing Systems Foundation, Inc.

Learning with auxiliary tasks can improve the ability of a primary task to generalise.However, this comes at the cost of manually labelling auxiliary data. We propose anew method which automatically learns appropriate labels for an auxiliary task,such that any supervised learning task can be improved without requiring access toany further data. The approach is to train two neural networks: a label-generationnetwork to predict the auxiliary labels, and a multi-task network to train theprimary task alongside the auxiliary task. The loss for the label-generation networkincorporates the loss of the multi-task network, and so this interaction between thetwo networks can be seen as a form of meta learning with a double gradient. Weshow that our proposed method, Meta AuXiliary Learning (MAXL), outperformssingle-task learning on 7 image datasets, without requiring any additional data.We also show that MAXL outperforms several other baselines for generatingauxiliary labels, and is even competitive when compared with human-definedauxiliary labels. The self-supervised nature of our method leads to a promisingnew direction towards automated generalisation. Source code can be found athttps://github.com/lorenmt/maxl.

Conference paper

Saeedi S, Carvalho EDC, Li W, Tzoumanikas D, Leutenegger S, Kelly PHJ, Davison AJet al., 2019, Characterizing visual localization and mapping datasets, 2019 International Conference on Robotics and Automation (ICRA), Publisher: Institute of Electrical and Electronics Engineers, ISSN: 1050-4729

Benchmarking mapping and motion estimation algorithms is established practice in robotics and computer vision. As the diversity of datasets increases, in terms of the trajectories, models, and scenes, it becomes a challenge to select datasets for a given benchmarking purpose. Inspired by the Wasserstein distance, this paper addresses this concern by developing novel metrics to evaluate trajectories and the environments without relying on any SLAM or motion estimation algorithm. The metrics, which so far have been missing in the research community, can be applied to the plethora of datasets that exist. Additionally, to improve the robotics SLAM benchmarking, the paper presents a new dataset for visual localization and mapping algorithms. A broad range of real-world trajectories is used in very high-quality scenes and a rendering framework to create a set of synthetic datasets with ground-truth trajectory and dense map which are representative of key SLAM applications such as virtual reality (VR), micro aerial vehicle (MAV) flight, and ground robotics.

Conference paper

Bujanca M, Gafton P, Saeedi S, Nisbet A, Bodin B, O'Boyle MFP, Davison AJ, Paul HJ K, Riley G, Lennox B, Lujan M, Furber Set al., 2019, SLAMBench 3.0: Systematic automated reproducible evaluation of SLAM systems for robot vision challenges and scene understanding, 2019 International Conference on Robotics and Automation (ICRA), Publisher: Institute of Electrical and Electronics Engineers, ISSN: 1050-4729

As the SLAM research area matures and the number of SLAM systems available increases, the need for frameworks that can objectively evaluate them against prior work grows. This new version of SLAMBench moves beyond traditional visual SLAM, and provides new support for scene understanding and non-rigid environments (dynamic SLAM). More concretely for dynamic SLAM, SLAMBench 3.0 includes the first publicly available implementation of DynamicFusion, along with an evaluation infrastructure. In addition, we include two SLAM systems (one dense, one sparse) augmented with convolutional neural networks for scene understanding, together with datasets and appropriate metrics. Through a series of use-cases, we demonstrate the newly incorporated algorithms, visulation aids and metrics (6 new metrics, 4 new datasets and 5 new algorithms).

Conference paper

Xu B, Li W, Tzoumanikas D, Bloesch M, Davison A, Leutenegger Set al., 2019, MID-fusion: octree-based object-level multi-instance dynamic SLAM, ICRA 2019- IEEE International Conference on Robotics and Automation, Publisher: IEEE

We propose a new multi-instance dynamic RGB-D SLAM system using anobject-level octree-based volumetric representation. It can provide robustcamera tracking in dynamic environments and at the same time, continuouslyestimate geometric, semantic, and motion properties for arbitrary objects inthe scene. For each incoming frame, we perform instance segmentation to detectobjects and refine mask boundaries using geometric and motion information.Meanwhile, we estimate the pose of each existing moving object using anobject-oriented tracking method and robustly track the camera pose against thestatic scene. Based on the estimated camera pose and object poses, we associatesegmented masks with existing models and incrementally fuse correspondingcolour, depth, semantic, and foreground object probabilities into each objectmodel. In contrast to existing approaches, our system is the first system togenerate an object-level dynamic volumetric map from a single RGB-D camera,which can be used directly for robotic tasks. Our method can run at 2-3 Hz on aCPU, excluding the instance segmentation part. We demonstrate its effectivenessby quantitatively and qualitatively testing it on both synthetic and real-worldsequences.

Conference paper

Zhi S, Bloesch M, Leutenegger S, Davison AJet al., 2019, SceneCode: Monocular Dense Semantic Reconstruction using Learned Encoded Scene Representations, Publisher: IEEE

Working paper

Bloesch M, Czarnowski J, Clark R, Leutenegger S, Davison AJet al., 2018, CodeSLAM - Learning a compact, optimisable representation for dense visual SLAM, IEEE Computer Vision and Pattern Recognition 2018, Publisher: IEEE, Pages: 2560-2568

The representation of geometry in real-time 3D perception systems continues to be a critical research issue. Dense maps capture complete surface shape and can be augmented with semantic labels, but their high dimensionality makes them computationally costly to store and process, and unsuitable for rigorous probabilistic inference. Sparse feature-based representations avoid these problems, but capture only partial scene information and are mainly useful for localisation only. We present a new compact but dense representation of scene geometry which is conditioned on the intensity data from a single image and generated from a code consisting of a small number of parameters. We are inspired by work both on learned depth from images, and auto-encoders. Our approach is suitable for use in a keyframe-based monocular dense SLAM system: While each keyframe with a code can produce a depth map, the code can be optimised efficiently jointly with pose variables and together with the codes of overlapping keyframes to attain global consistency. Conditioning the depth map on the image allows the code to only represent aspects of the local geometry which cannot directly be predicted from the image. We explain how to learn our code representation, and demonstrate its advantageous properties in monocular SLAM.

Conference paper

Saeedi Gharahbolagh S, Bodin B, Wagstaff H, Nisbet A, Nardi L, Mawer J, Melot N, Palomar O, Vespa E, Gorgovan C, Webb A, Clarkson J, Tomusk E, Debrunner T, Kaszyk K, Gonzalez P, Rodchenko A, Riley G, Kotselidis C, Franke B, OBoyle M, Davison A, Kelly P, Lujan M, Furber Set al., 2018, Navigating the landscape for real-time localisation and mapping for robotics, virtual and augmented reality, Proceedings of the IEEE, Vol: 106, Pages: 2020-2039, ISSN: 0018-9219

Visual understanding of 3-D environments in real time, at low power, is a huge computational challenge. Often referred to as simultaneous localization and mapping (SLAM), it is central to applications spanning domestic and industrial robotics, autonomous vehicles, and virtual and augmented reality. This paper describes the results of a major research effort to assemble the algorithms, architectures, tools, and systems software needed to enable delivery of SLAM, by supporting applications specialists in selecting and configuring the appropriate algorithm and the appropriate hardware, and compilation pathway, to meet their performance, accuracy, and energy consumption goals. The major contributions we present are: 1) tools and methodology for systematic quantitative evaluation of SLAM algorithms; 2) automated, machine-learning-guided exploration of the algorithmic and implementation design space with respect to multiple objectives; 3) end-to-end simulation tools to enable optimization of heterogeneous, accelerated architectures for the specific algorithmic requirements of the various SLAM algorithmic approaches; and 4) tools for delivering, where appropriate, accelerated, adaptive SLAM solutions in a managed, JIT-compiled, adaptive runtime context.

Journal article

Matas J, James S, Davison A, 2018, Sim-to-real reinforcement learning for deformable object manipulation, Conference on Robot Learning 2018, Publisher: PMLR, Pages: 734-743

We have seen much recent progress in rigid object manipulation, but in-teraction with deformable objects has notably lagged behind. Due to the large con-figuration space of deformable objects, solutions using traditional modelling ap-proaches require significant engineering work. Perhaps then, bypassing the needfor explicit modelling and instead learning the control in an end-to-end mannerserves as a better approach? Despite the growing interest in the use of end-to-endrobot learning approaches, only a small amount of work has focused on their ap-plicability to deformable object manipulation. Moreover, due to the large amountof data needed to learn these end-to-end solutions, an emerging trend is to learncontrol policies in simulation and then transfer them over to the real world. To-date, no work has explored whether it is possible to learn and transfer deformableobject policies. We believe that if sim-to-real methods are to be employed fur-ther, then it should be possible to learn to interact with a wide variety of objects,and not only rigid objects. In this work, we use a combination of state-of-the-artdeep reinforcement learning algorithms to solve the problem of manipulating de-formable objects (specifically cloth). We evaluate our approach on three tasks —folding a towel up to a mark, folding a face towel diagonally, and draping a pieceof cloth over a hanger. Our agents are fully trained in simulation with domainrandomisation, and then successfully deployed in the real world without havingseen any real deformable objects.

Conference paper

James S, Blosch M, Davison A, 2018, Task-embedded control networks for few-shot imitation learning, Conference on Robot Learning, Publisher: PMLR, Pages: 783-795

Much like humans, robots should have the ability to leverage knowledge from previously learned tasks in order to learn new tasks quickly in new and unfamiliar environments. Despite this, most robot learning approaches have focused on learning a single task, from scratch, with a limited notion of generalisation, and no way of leveraging the knowledge to learn other tasks more efficiently. One possible solution is meta-learning, but many of the related approaches are limited in their ability to scale to a large number of tasks and to learn further tasks without forgetting previously learned ones. With this in mind, we introduce Task-Embedded Control Networks, which employ ideas from metric learning in order to create a task embedding that can be used by a robot to learn new tasks from one or more demonstrations. In the area of visually-guided manipulation, we present simulation results in which we surpass the performance of a state-of-the-art method when using only visual information from each demonstration. Additionally, we demonstrate that our approach can also be used in conjunction with domain randomisation to train our few-shot learning ability in simulation and then deploy in the real world without any additional training. Once deployed, the robot can learn new tasks from a single real-world demonstration.

Conference paper

This data is extracted from the Web of Science and reproduced under a licence from Thomson Reuters. You may not copy or re-distribute this data in whole or in part without the written consent of the Science business of Thomson Reuters.

Request URL: http://wlsprd.imperial.ac.uk:80/respub/WEB-INF/jsp/search-html.jsp Request URI: /respub/WEB-INF/jsp/search-html.jsp Query String: respub-action=search.html&id=00450245&limit=30&person=true