Search

This study explores the potential of off-the-shelf Vision-Language Models (VLMs) for high-level robot planning in the context of autonomous navigation. Indeed, while most of existing learning-based approaches for path planning require extensive task-specific training/fine-tuning, we demonstrate how such training can be avoided for most practical cases. To do this, we introduce Select2Plan (S2P), a novel training-free framework for high-level robot planning which completely eliminates the need for fine-tuning or specialised training. By leveraging structured Visual Question-Answering (VQA) and In-Context Learning (ICL), our approach drastically reduces the need for data collection, requiring a fraction of the task-specific data typically used by trained models, or even relying only on online data. Our method facilitates the effective use of a generally trained VLM in a flexible and cost-efficient way, and does not require additional sensing except for a simple monocular camera. We demonstrate its adaptability across various scene types, context sources, and sensing setups. We evaluate our approach in two distinct scenarios: traditional First-Person View (FPV) and infrastructure-driven Third-Person View (TPV) navigation, demonstrating the flexibility and simplicity of our method. Our technique significantly enhances the navigational capabilities of a baseline VLM of approximately 50% in TPV scenario, and is comparable to trained models in the FPV one, with as few as 20 demonstrations.

In recent years, the serverless paradigm has been widely adopted to develop cloud applications, as it enables building scalable solutions while delegating operational concerns such as infrastructure management and resource provisioning to the serverless provider. Despite bringing undisputed advantages, the serverless model requires a change in programming paradigm that may add complexity in software development. In particular, in the Function-as-a-Service (FaaS) paradigm, functions are inherently stateless. As a consequence, developers carry the burden of directly interacting with external storage services and handling concurrency and state consistency across function invocations. This results in less time spent on solving the actual business problems they face. Moving from these premises, this paper proposes Histrio, a programming model and execution environment that simplifies the development of complex stateful applications in the FaaS paradigm. Histrio grounds on the actor programming model, and lifts concerns such as state management, database interaction, and concurrency handling from developers. It enriches the actor model with features that simplify and optimize the interaction with external storage. It guarantees exactly-once-processing consistency, meaning that the application always behaves as if any interaction with external clients was processed once and only once, masking failures. Histrio has been compared with a classical FaaS implementation to evaluate both the development time saved due to the guarantees the system offers and the applicability of Histrio in typical applications. In the evaluated scenarios, Histrio simplified the implementation by significantly removing the amount of code needed to handle operational concerns. It proves to be scalable and it provides configuration mechanisms to trade performance and execution costs.

This paper proposes a method for on-demand scenario generation in simulation, grounded on real-world data. Evaluating the behaviour of Autonomous Vehicles (AVs) in both safety-critical and regular scenarios is essential for assessing their robustness before real-world deployment. By integrating scenarios derived from real-world datasets into the simulation, we enhance the plausibility and validity of testing sets. This work introduces a novel approach that employs temporal scene graphs to capture evolving spatiotemporal relationships among scene entities from a real-world dataset, enabling the generation of dynamic scenarios in simulation through Graph Neural Networks (GNNs). User-defined action and criticality conditioning are used to ensure flexible, tailored scenario creation. Our model significantly outperforms the benchmarks in accurately predicting links corresponding to the requested scenarios. We further evaluate the validity and compatibility of our generated scenarios in an off-the-shelf simulator., Comment: 7 pages, 8 figures

This study explores the intersection of neural networks and classical robotics algorithms through the Neural Algorithmic Reasoning (NAR) framework, allowing to train neural networks to effectively reason like classical robotics algorithms by learning to execute them. Algorithms are integral to robotics and safety-critical applications due to their predictable and consistent performance through logical and mathematical principles. In contrast, while neural networks are highly adaptable, handling complex, high-dimensional data and generalising across tasks, they often lack interpretability and transparency in their internal computations. We propose a Graph Neural Network (GNN)-based learning framework, NAR-*ICP, which learns the intermediate algorithmic steps of classical ICP-based pointcloud registration algorithms, and extend the CLRS Algorithmic Reasoning Benchmark with classical robotics perception algorithms. We evaluate our approach across diverse datasets, from real-world to synthetic, demonstrating its flexibility in handling complex and noisy inputs, along with its potential to be used as part of a larger learning system. Our results indicate that our method achieves superior performance across all benchmarks and datasets, consistently surpassing even the algorithms it has been trained on, further demonstrating its ability to generalise beyond the capabilities of traditional algorithms., Comment: 17 pages, 9 figures

This paper investigates the utility of Neural Radiance Fields (NeRF) models in extending the regions of operation of a mobile robot, controlled by Image-Based Visual Servoing (IBVS) via static CCTV cameras. Using NeRF as a 3D-representation prior, the robot's footprint may be extrapolated geometrically and used to train a CNN-based network to extract it online from the robot's appearance alone. The resulting footprint results in a tighter bound than a robot-wide bounding box, allowing the robot's controller to prescribe more optimal trajectories and expand its safe operational floor area., Comment: Accepted as extended abstract for ICRA@40

We describe a challenging robotics deployment in a complex ecosystem to monitor a rich plant community. The study site is dominated by dynamic grassland vegetation and is thus visually ambiguous and liable to drastic appearance change over the course of a day and especially through the growing season. This dynamism and complexity in appearance seriously impact the stability of the robotics platform, as localisation is a foundational part of that control loop, and so routes must be carefully taught and retaught until autonomy is robust and repeatable. Our system is demonstrated over a 6-week period monitoring the response of grass species to experimental climate change manipulations. We also discuss the applicability of our pipeline to monitor biodiversity in other complex natural settings., Comment: to be presented at the Workshop on Field Robotics - ICRA 2024

This paper adapts a general dataset representation technique to produce robust Visual Place Recognition (VPR) descriptors, crucial to enable real-world mobile robot localisation. Two parallel lines of work on VPR have shown, on one side, that general-purpose off-the-shelf feature representations can provide robustness to domain shifts, and, on the other, that fused information from sequences of images improves performance. In our recent work on measuring domain gaps between image datasets, we proposed a Visual Distribution of Neuron Activations (VDNA) representation to represent datasets of images. This representation can naturally handle image sequences and provides a general and granular feature representation derived from a general-purpose model. Moreover, our representation is based on tracking neuron activation values over the list of images to represent and is not limited to a particular neural network layer, therefore having access to high- and low-level concepts. This work shows how VDNAs can be used for VPR by learning a very lightweight and simple encoder to generate task-specific descriptors. Our experiments show that our representation can allow for better robustness than current solutions to serious domain shifts away from the training data distribution, such as to indoor environments and aerial imagery., Comment: Published at ICRA 2024

This paper introduces RobotCycle, a novel ongoing project that leverages Autonomous Vehicle (AV) research to investigate how road infrastructure influences cyclist behaviour and safety during real-world journeys. The project's requirements were defined in collaboration with key stakeholders, including city planners, cyclists, and policymakers, informing the design of risk and safety metrics and the data collection criteria. We propose a data-driven approach relying on a novel, rich dataset of diverse traffic scenes and scenarios captured using a custom-designed wearable sensing unit. By analysing road-user trajectories, we identify normal path deviations indicating potential risks or hazardous interactions related to infrastructure elements in the environment. Our analysis correlates driving profiles and trajectory patterns with local road segments, driving conditions, and road-user interactions to predict traffic behaviours and identify critical scenarios. Moreover, by leveraging advancements in AV research, the project generates detailed 3D High-Definition Maps (HD Maps), traffic flow patterns, and trajectory models to provide a comprehensive assessment and analysis of the behaviour of all traffic agents. These data can then inform the design of cyclist-friendly road infrastructure, ultimately enhancing road safety and cyclability. The project provides valuable insights for enhancing cyclist protection and advancing sustainable urban mobility., Comment: IEEE Intelligent Vehicles Symposium (IV 2024)

This paper is about 3D pose estimation on LiDAR scans with extremely minimal storage requirements to enable scalable mapping and localisation. We achieve this by clustering all points of segmented scans into semantic objects and representing them only with their respective centroid and semantic class. In this way, each LiDAR scan is reduced to a compact collection of four-number vectors. This abstracts away important structural information from the scenes, which is crucial for traditional registration approaches. To mitigate this, we introduce an object-matching network based on self- and cross-correlation that captures geometric and semantic relationships between entities. The respective matches allow us to recover the relative transformation between scans through weighted Singular Value Decomposition (SVD) and RANdom SAmple Consensus (RANSAC). We demonstrate that such representation is sufficient for metric localisation by registering point clouds taken under different viewpoints on the KITTI dataset, and at different periods of time localising between KITTI and KITTI-360. We achieve accurate metric estimates comparable with state-of-the-art methods with almost half the representation size, specifically 1.33 kB on average., Comment: Accepted for publication at the IEEE International Conference on Robotics and Automation (ICRA) 2024

There is a growing academic interest as well as commercial exploitation of millimetre-wave scanning radar for autonomous vehicle localisation and scene understanding. Although several datasets to support this research area have been released, they are primarily focused on urban or semi-urban environments. Nevertheless, rugged offroad deployments are important application areas which also present unique challenges and opportunities for this sensor technology. Therefore, the Oxford Offroad Radar Dataset (OORD) presents data collected in the rugged Scottish highlands in extreme weather. The radar data we offer to the community are accompanied by GPS/INS reference - to further stimulate research in radar place recognition. In total we release over 90GiB of radar scans as well as GPS and IMU readings by driving a diverse set of four routes over 11 forays, totalling approximately 154km of rugged driving. This is an area increasingly explored in literature, and we therefore present and release examples of recent open-sourced radar place recognition systems and their performance on our dataset. This includes a learned neural network, the weights of which we also release. The data and tools are made freely available to the community at https://oxford-robotics-institute.github.io/oord-dataset.

Knowing when a trained segmentation model is encountering data that is different to its training data is important. Understanding and mitigating the effects of this play an important part in their application from a performance and assurance perspective - this being a safety concern in applications such as autonomous vehicles (AVs). This work presents a segmentation network that can detect errors caused by challenging test domains without any additional annotation in a single forward pass. As annotation costs limit the diversity of labelled datasets, we use easy-to-obtain, uncurated and unlabelled data to learn to perform uncertainty estimation by selectively enforcing consistency over data augmentation. To this end, a novel segmentation benchmark based on the SAX Dataset is used, which includes labelled test data spanning three autonomous-driving domains, ranging in appearance from dense urban to off-road. The proposed method, named Gamma-SSL, consistently outperforms uncertainty estimation and Out-of-Distribution (OoD) techniques on this difficult benchmark - by up to 10.7% in area under the receiver operating characteristic (ROC) curve and 19.2% in area under the precision-recall (PR) curve in the most challenging of the three scenarios., Comment: Accepted for publication in IEEE Transactions on Robotics (T-RO)

This work proposes a semantic segmentation network that produces high-quality uncertainty estimates in a single forward pass. We exploit general representations from foundation models and unlabelled datasets through a Masked Image Modeling (MIM) approach, which is robust to augmentation hyper-parameters and simpler than previous techniques. For neural networks used in safety-critical applications, bias in the training data can lead to errors; therefore it is crucial to understand a network's limitations at run time and act accordingly. To this end, we test our proposed method on a number of test domains including the SAX Segmentation benchmark, which includes labelled test data from dense urban, rural and off-road driving domains. The proposed method consistently outperforms uncertainty estimation and Out-of-Distribution (OoD) techniques on this difficult benchmark., Comment: Accepted for publication at 2024 IEEE International Conference on Robotics and Automation (ICRA)

We present a system which grows and manages a network of remote viewpoints during the natural installation cycle for a newly installed camera network or a newly deployed robot fleet. No explicit notion of camera position or orientation is required, neither global - i.e. relative to a building plan - nor local - i.e. relative to an interesting point in a room. Furthermore, no metric relationship between viewpoints is required. Instead, we leverage our prior work in effective remote control without extrinsic or intrinsic calibration and extend it to the multi-camera setting. In this, we memorise, from simultaneous robot detections in the tracker thread, soft pixel-wise topological connections between viewpoints. We demonstrate our system with repeated autonomous traversals of workspaces connected by a network of six cameras across a productive office environment., Comment: Paper accepted to the 18th International Symposium on Experimental Robotics (ISER 2023)

Recalling the most relevant visual memories for localisation or understanding a priori the likely outcome of localisation effort against a particular visual memory is useful for efficient and robust visual navigation. Solutions to this problem should be divorced from performance appraisal against ground truth - as this is not available at run-time - and should ideally be based on generalisable environmental observations. For this, we propose applying the recently developed Visual DNA as a highly scalable tool for comparing datasets of images - in this work, sequences of map and live experiences. In the case of localisation, important dataset differences impacting performance are modes of appearance change, including weather, lighting, and season. Specifically, for any deep architecture which is used for place recognition by matching feature volumes at a particular layer, we use distribution measures to compare neuron-wise activation statistics between live images and multiple previously recorded past experiences, with a potentially large seasonal (winter/summer) or time of day (day/night) shift. We find that differences in these statistics correlate to performance when localising using a past experience with the same appearance gap. We validate our approach over the Nordland cross-season dataset as well as data from Oxford's University Parks with lighting and mild seasonal change, showing excellent ability of our system to rank actual localisation performance across candidate experiences., Comment: 18th International Symposium on Experimental Robotics (ISER 2023)

In this paper, we propose a robot oriented knowledge management system based on the use of the Prolog language. Our framework hinges on a special organisation of knowledge base that enables: 1. its efficient population from natural language texts using semi-automated procedures based on Large Language Models, 2. the bumpless generation of temporal parallel plans for multi-robot systems through a sequence of transformations, 3. the automated translation of the plan into an executable formalism (the behaviour trees). The framework is supported by a set of open source tools and is shown on a realistic application., Comment: 7 pages, 4 figures, submitted to ICRA 2024

This work explores Doppler information from a millimetre-Wave (mm-W) Frequency-Modulated Continuous-Wave (FMCW) scanning radar to make odometry estimation more robust and accurate. Firstly, doppler information is added to the scan masking process to enhance correlative scan matching. Secondly, we train a Neural Network (NN) for regressing forward velocity directly from a single radar scan; we fuse this estimate with the correlative scan matching estimate and show improved robustness to bad estimates caused by challenging environment geometries, e.g. narrow tunnels. We test our method with a novel custom dataset which is released with this work at https://ori.ox.ac.uk/publications/datasets., Comment: Accepted to ITSC 2023

In this work, we propose a methodology for investigating the use of semantic attention to enhance the explainability of Graph Neural Network (GNN)-based models. Graph Deep Learning (GDL) has emerged as a promising field for tasks like scene interpretation, leveraging flexible graph structures to concisely describe complex features and relationships. As traditional explainability methods used in eXplainable AI (XAI) cannot be directly applied to such structures, graph-specific approaches are introduced. Attention has been previously employed to estimate the importance of input features in GDL, however, the fidelity of this method in generating accurate and consistent explanations has been questioned. To evaluate the validity of using attention weights as feature importance indicators, we introduce semantically-informed perturbations and correlate predicted attention weights with the accuracy of the model. Our work extends existing attention-based graph explainability methods by analysing the divergence in the attention distributions in relation to semantically sorted feature sets and the behaviour of a GNN model, efficiently estimating feature importance. We apply our methodology on a lidar pointcloud estimation model successfully identifying key semantic classes that contribute to enhanced performance, effectively generating reliable post-hoc semantic explanations., Comment: International Conference on Advanced Robotics (ICAR 2023)

This paper proposes a Graph Neural Network(GNN)-based method for exploiting semantics and local geometry to guide the identification of reliable pointcloud registration candidates. Semantic and morphological features of the environment serve as key reference points for registration, enabling accurate lidar-based pose estimation. Our novel lightweight static graph structure informs our attention-based node aggregation network by identifying semantic-instance relationships, acting as an inductive bias to significantly reduce the computational burden of pointcloud registration. By connecting candidate nodes and exploiting cross-graph attention, we identify confidence scores for all potential registration correspondences and estimate the displacement between pointcloud scans. Our pipeline enables introspective analysis of the model's performance by correlating it with the individual contributions of local structures in the environment, providing valuable insights into the system's behaviour. We test our method on the KITTI odometry dataset, achieving competitive accuracy compared to benchmark methods and a higher track smoothness while relying on significantly fewer network parameters., Comment: International Conference on Advanced Robotics (ICAR 2023)

This work proposes a fast deployment pipeline for visually-servoed robots which does not assume anything about either the robot - e.g. sizes, colour or the presence of markers - or the deployment environment. In this, accurate estimation of robot orientation is crucial for successful navigation in complex environments; manual labelling of angular values is, though, time-consuming and possibly hard to perform. For this reason, we propose a weakly supervised pipeline that can produce a vast amount of data in a small amount of time. We evaluate our approach on a dataset of remote camera images captured in various indoor environments demonstrating high tracking performances when integrated into a fully-autonomous pipeline with a simple controller. With this, we then analyse the data requirement of our approach, showing how it is possible to deploy a new robot in a new environment in less than 30.00 min., Comment: Accepted at IROS 2023

Today, data analysis drives the decision-making process in virtually every human activity. This demands for software platforms that offer simple programming abstractions to express data analysis tasks and that can execute them in an efficient and scalable way. State-of-the-art solutions range from low-level programming primitives, which give control to the developer about communication and resource usage, but require significant effort to develop and optimize new algorithms, to high-level platforms that hide most of the complexities of parallel and distributed processing, but often at the cost of reduced efficiency. To reconcile these requirements, we developed Renoir, a novel distributed data processing platform written in Rust. Renoir provides a high-level dataflow programming model as mainstream data processing systems. It supports static and streaming data, it enables data transformations, grouping, aggregation, iterative computations, and time-based analytics, incurring in a low overhead. This paper presents In this paper, we present the programming model and the implementation details of Renoir. We evaluate it under heterogeneous workloads. We compare it with state-of-the-art solutions for data analysis and high-performance computing, as well as alternative research products, which offer different programming abstractions and implementation strategies. Renoir programs are compact and easy to write: developers need not care about low-level concerns such as resource usage, data serialization, concurrency control, and communication. Renoir consistently presents comparable or better performance than competing solutions, by a large margin in several scenarios. We conclude that Renoir offers a good tradeoff between simplicity and performance, allowing developers to easily express complex data analysis tasks and achieve high performance and scalability., Comment: Submitted to Future Generation Computer Systems, December 2023

Recent advances in the field of very long distance optical communication suggest the adoption of the advanced technology based on Hollow Core Nested Anti-resonant Nodeless Fiber (HC-NANF) within the endeavour of Gravitational Wave detection using a Mach-Zehnder optical interferometer (MZ-IF). The proposal, consisting of a summary project of the device emphasizes the favorable properties of (MZ-IF) in comparison with Michelson Interferometer (MIF) currently in operation. The key feature of the proposed method consists of the use of a couple of "fibrated" metallic antennas enfolded by a very large (K x 8.10^4 with K=1,2,3 etc.) of coiled (HC-NANF) rings. This amounts to a corresponding fiber length: Leff = K x 1600 Km. The relevant properties of the device are noise reduction, absence of critical optical mirror alignment in a noisy environment, reduced spatial extension of the apparatus, exploration of the entire sky scenario by freely orientable antennas, a substational cost reduction of the apparatus. The remarkable properties of (HC-NANF), invented by F. Poletti in 2013 are currently investigated by his group at the University of Southampton (UK)., Comment: 10 pages, 10 figures

This paper proposes a simple strategy for sim-to-real in Deep-Reinforcement Learning (DRL) -- called Roll-Drop -- that uses dropout during simulation to account for observation noise during deployment without explicitly modelling its distribution for each state. DRL is a promising approach to control robots for highly dynamic and feedback-based manoeuvres, and accurate simulators are crucial to providing cheap and abundant data to learn the desired behaviour. Nevertheless, the simulated data are noiseless and generally show a distributional shift that challenges the deployment on real machines where sensor readings are affected by noise. The standard solution is modelling the latter and injecting it during training; while this requires a thorough system identification, Roll-Drop enhances the robustness to sensor noise by tuning only a single parameter. We demonstrate an 80% success rate when up to 25% noise is injected in the observations, with twice higher robustness than the baselines. We deploy the controller trained in simulation on a Unitree A1 platform and assess this improved robustness on the physical system., Comment: Accepted at Learning for Dynamics & Control Conference 2023 (L4DC), 10 pages, 7 figures

Selecting appropriate datasets is critical in modern computer vision. However, no general-purpose tools exist to evaluate the extent to which two datasets differ. For this, we propose representing images - and by extension datasets - using Distributions of Neuron Activations (DNAs). DNAs fit distributions, such as histograms or Gaussians, to activations of neurons in a pre-trained feature extractor through which we pass the image(s) to represent. This extractor is frozen for all datasets, and we rely on its generally expressive power in feature space. By comparing two DNAs, we can evaluate the extent to which two datasets differ with granular control over the comparison attributes of interest, providing the ability to customise the way distances are measured to suit the requirements of the task at hand. Furthermore, DNAs are compact, representing datasets of any size with less than 15 megabytes. We demonstrate the value of DNAs by evaluating their applicability on several tasks, including conditional dataset comparison, synthetic image evaluation, and transfer learning, and across diverse datasets, ranging from synthetic cat images to celebrity faces and urban driving scenes., Comment: Published at CVPR 2023. Project page with code: https://bramtoula.github.io/vdna/

This work explores scene graphs as a distilled representation of high-level information for autonomous driving, applied to future driver-action prediction. Given the scarcity and strong imbalance of data samples, we propose a self-supervision pipeline to infer representative and well-separated embeddings. Key aspects are interpretability and explainability; as such, we embed in our architecture attention mechanisms that can create spatial and temporal heatmaps on the scene graphs. We evaluate our system on the ROAD dataset against a fully-supervised approach, showing the superiority of our training regime., Comment: Accepted to the 2023 IEEE International Conference on Robotics and Automation (ICRA)

The metaverse has the potential to revolutionize the next generation of the Internet by supporting highly interactive services with the help of Mixed Reality (MR) technologies; still, to provide a satisfactory experience for users, the synchronization between the physical world and its digital models is crucial. This work proposes a sampling, communication and prediction co-design framework to minimize the communication load subject to a constraint on tracking the Mean Squared Error (MSE) between a real-world device and its digital model in the metaverse. To optimize the sampling rate and the prediction horizon, we exploit expert knowledge and develop a constrained Deep Reinforcement Learning (DRL) algorithm, named Knowledge-assisted Constrained Twin-Delayed Deep Deterministic (KC-TD3) policy gradient algorithm. We validate our framework on a prototype composed of a real-world robotic arm and its digital model. Compared with existing approaches: (1) When the tracking error constraint is stringent (MSE=0.002 degrees), our policy degenerates into the policy in the sampling-communication co-design framework. (2) When the tracking error constraint is mild (MSE=0.007 degrees), our policy degenerates into the policy in the prediction-communication co-design framework. (3) Our framework achieves a better trade-off between the average MSE and the average communication load compared with a communication system without sampling and prediction. For example, the average communication load can be reduced up to 87% when the track error constraint is 0.002 degrees. (4) Our policy outperforms the benchmark with the static sampling rate and prediction horizon optimized by exhaustive search, in terms of the tail probability of the tracking error. Furthermore, with the assistance of expert knowledge, the proposed algorithm KC-TD3 achieves better convergence time, stability, and final policy performance.

This paper is about extremely robust and lightweight localisation using LiDAR point clouds based on instance segmentation and graph matching. We model 3D point clouds as fully-connected graphs of semantically identified components where each vertex corresponds to an object instance and encodes its shape. Optimal vertex association across graphs allows for full 6-Degree-of-Freedom (DoF) pose estimation and place recognition by measuring similarity. This representation is very concise, condensing the size of maps by a factor of 25 against the state-of-the-art, requiring only 3kB to represent a 1.4MB laser scan. We verify the efficacy of our system on the SemanticKITTI dataset, where we achieve a new state-of-the-art in place recognition, with an average of 88.4% recall at 100% precision where the next closest competitor follows with 64.9%. We also show accurate metric pose estimation performance - estimating 6-DoF pose with median errors of 10 cm and 0.33 deg., Comment: Accepted for publication at the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) 2022

This paper presents a method that leverages vehicle motion constraints to refine data associations in a point-based radar odometry system. By using the strong prior on how a non-holonomic robot is constrained to move smoothly through its environment, we develop the necessary framework to estimate ego-motion from a single landmark association rather than considering all of these correspondences at once. This allows for informed outlier detection of poor matches that are a dominant source of pose estimate error. By refining the subset of matched landmarks, we see an absolute decrease of 2.15% (from 4.68% to 2.53%) in translational error, approximately halving the error in odometry (reducing by 45.94%) than when using the full set of correspondences. This contribution is relevant to other point-based odometry implementations that rely on a range sensor and provides a lightweight and interpretable means of incorporating vehicle dynamics for ego-motion estimation., Comment: Accepted for RA-L

In this paper we present a compositing image synthesis method that generates RGB canvases with well aligned segmentation maps and sparse depth maps, coupled with an in-painting network that transforms the RGB canvases into high quality RGB images and the sparse depth maps into pixel-wise dense depth maps. We benchmark our method in terms of structural alignment and image quality, showing an increase in mIoU over SOTA by 3.7 percentage points and a highly competitive FID. Furthermore, we analyse the quality of the generated data as training data for semantic segmentation and depth completion, and show that our approach is more suited for this purpose than other methods.