We infer and generate three-dimensional (3D) scene information from a single input image and without supervision. This problem is under-explored, with most prior work relying on supervision from, e.g., 3D ground-truth, multiple images of a scene, image silhouettes or key-points. We propose Pix2Shape, an approach to solve this problem with four components: (i) an encoder that infers the latent 3D representation from an image, (ii) a decoder that generates an explicit 2.5D surfel-based reconstruction of a scene from the latent code (iii) a differentiable renderer that synthesizes a 2D image from the surfel representation, and (iv) a critic network trained to discriminate between images generated by the decoder-renderer and those from a training distribution. Pix2Shape can generate complex 3D scenes that scale with the view-dependent on-screen resolution, unlike representations that capture world-space resolution, i.e., voxels or meshes. We show that Pix2Shape learns a consistent scene representation in its encoded latent space and that the decoder can then be applied to this latent representation in order to synthesize the scene from a novel viewpoint. We evaluate Pix2Shape with experiments on the ShapeNet dataset as well as on a novel benchmark we developed, called 3D-IQTT, to evaluate models based on their ability to enable 3d spatial reasoning. Qualitative and quantitative evaluation demonstrate Pix2Shape's ability to solve scene reconstruction, generation, and understanding tasks.
Spatial reasoning is an important component of human intelligence. We can imagine the shapes of 3D objects and reason about their spatial relations by merely looking at their three-view line drawings in 2D, with different levels of competence. Can deep networks be trained to perform spatial reasoning tasks? How can we measure their "spatial intelligence"? To answer these questions, we present the SPARE3D dataset. Based on cognitive science and psychometrics, SPARE3D contains three types of 2D-3D reasoning tasks on view consistency, camera pose, and shape generation, with increasing difficulty. We then design a method to automatically generate a large number of challenging questions with ground truth answers for each task. They are used to provide supervision for training our baseline models using state-of-the-art architectures like ResNet. Our experiments show that although convolutional networks have achieved superhuman performance in many visual learning tasks, their spatial reasoning performance on SPARE3D tasks is either lower than average human performance or even close to random guesses. We hope SPARE3D can stimulate new problem formulations and network designs for spatial reasoning to empower intelligent robots to operate effectively in the 3D world via 2D sensors. The dataset and code are available at https://ai4ce.github.io/SPARE3D.
Comprehensive visual understanding requires detection frameworks that can effectively learn and utilize object interactions while analyzing objects individually. This is the main objective in Human-Object Interaction (HOI) detection task. In particular, relative spatial reasoning and structural connections between objects are essential cues for analyzing interactions, which is addressed by the proposed Visual-Spatial-Graph Network (VSGNet) architecture. VSGNet extracts visual features from the human-object pairs, refines the features with spatial configurations of the pair, and utilizes the structural connections between the pair via graph convolutions. The performance of VSGNet is thoroughly evaluated using the Verbs in COCO (V-COCO) and HICO-DET datasets. Experimental results indicate that VSGNet outperforms state-of-the-art solutions by 8% or 4 mAP in V-COCO and 16% or 3 mAP in HICO-DET.
While traditional methods for instruction-following typically assume prior linguistic and perceptual knowledge, many recent works in reinforcement learning (RL) have proposed learning policies end-to-end, typically by training neural networks to map joint representations of observations and instructions directly to actions. In this work, we present a novel framework for learning to perform temporally extended tasks using spatial reasoning in the RL framework, by sequentially imagining visual goals and choosing appropriate actions to fulfill imagined goals. Our framework operates on raw pixel images, assumes no prior linguistic or perceptual knowledge, and learns via intrinsic motivation and a single extrinsic reward signal measuring task completion. We validate our method in two environments with a robot arm in a simulated interactive 3D environment. Our method outperforms two flat architectures with raw-pixel and ground-truth states, and a hierarchical architecture with ground-truth states on object arrangement tasks.
The ability to persist in the spacial environment is, not only in the robotic context, an essential feature. Positional knowledge is one of the most important aspects of space and a number of methods to represent these information have been developed in the in the research area of spatial cognition. The basic qualitative spatial representation and reasoning techniques are presented in this thesis and several calculi are briefly reviewed. Features and applications of qualitative calculi are summarized. A new calculus for representing and reasoning about qualitative spatial orientation and distances is being designed. It supports an arbitrary level of granularity over ternary relations of points. Ways of improving the complexity of the composition are shown and an implementation of the calculus demonstrates its capabilities. Existing qualitative spatial calculi of positional information are compared to the new approach and possibilities for future research are outlined.
The blocks world is a classic toy domain that has long been used to build and test spatial reasoning systems. Despite its relative simplicity, tackling this domain in its full complexity requires the agent to exhibit a rich set of functional capabilities, ranging from vision to natural language understanding. There is currently a resurgence of interest in solving problems in such limited domains using modern techniques. In this work we tackle spatial question answering in a holistic way, using a vision system, speech input and output mediated by an animated avatar, a dialogue system that robustly interprets spatial queries, and a constraint solver that derives answers based on 3-D spatial modeling. The contributions of this work include a semantic parser that maps spatial questions into logical forms consistent with a general approach to meaning representation, a dialog manager based on a schema representation, and a constraint solver for spatial questions that provides answers in agreement with human perception. These and other components are integrated into a multi-modal human-computer interaction pipeline.
Visual question answering (VQA) comprises a variety of language capabilities. The diagnostic benchmark dataset CLEVR has fueled progress by helping to better assess and distinguish models in basic abilities like counting, comparing and spatial reasoning in vitro. Following this approach, we focus on spatial language capabilities and investigate the question: what are the key ingredients to handle simple visual-spatial relations? We look at the SAN, RelNet, FiLM and MC models and evaluate their learning behavior on diagnostic data which is solely focused on spatial relations. Via comparative analysis and targeted model modification we identify what really is required to substantially improve upon the CNN-LSTM baseline.
Given the success of the deep convolutional neural networks (DCNNs) in applications of visual recognition and classification, it would be tantalizing to test if DCNNs can also learn spatial concepts, such as straightness, convexity, left/right, front/back, relative size, aspect ratio, polygons, etc., from varied visual examples of these concepts that are simple and yet vital for spatial reasoning. Much to our dismay, extensive experiments of the type of cognitive psychology demonstrate that the data-driven deep learning (DL) cannot see through superficial variations in visual representations and grasp the spatial concept in abstraction. The root cause of failure turns out to be the learning methodology, not the computational model of the neural network itself. By incorporating task-specific convolutional kernels, we are able to construct DCNNs for spatial cognition tasks that can generalize to input images not drawn from the same distribution of the training set. This work raises a precaution that without manually-incorporated priors or features DCCNs may fail spatial cognitive tasks at rudimentary level.
Understanding the spatial relations between objects in images is a surprisingly challenging task. A chair may be "behind" a person even if it appears to the left of the person in the image (depending on which way the person is facing). Two students that appear close to each other in the image may not in fact be "next to" each other if there is a third student between them. We introduce SpatialSense, a dataset specializing in spatial relation recognition which captures a broad spectrum of such challenges, allowing for proper benchmarking of computer vision techniques. SpatialSense is constructed through adversarial crowdsourcing, in which human annotators are tasked with finding spatial relations that are difficult to predict using simple cues such as 2D spatial configuration or language priors. Adversarial crowdsourcing significantly reduces dataset bias and samples more interesting relations in the long tail compared to existing datasets. On SpatialSense, state-of-the-art recognition models perform comparably to simple baselines, suggesting that they rely on straightforward cues instead of fully reasoning about this complex task. The SpatialSense benchmark provides a path forward to advancing the spatial reasoning capabilities of computer vision systems. The dataset and code are available at https://github.com/princeton-vl/SpatialSense.
In this paper, we argue that simulation platforms enable a novel type of embodied spatial reasoning, one facilitated by a formal model of object and event semantics that renders the continuous quantitative search space of an open-world, real-time environment tractable. We provide examples for how a semantically-informed AI system can exploit the precise, numerical information provided by a game engine to perform qualitative reasoning about objects and events, facilitate learning novel concepts from data, and communicate with a human to improve its models and demonstrate its understanding. We argue that simulation environments, and game engines in particular, bring together many different notions of "simulation" and many different technologies to provide a highly-effective platform for developing both AI systems and tools to experiment in both machine and human intelligence.
This book contains the accepted papers at ProSocrates 2017 Symposium: Problem-solving,Creativity and Spatial Reasoning in Cognitive Systems. ProSocrates 2017 symposium was held at the Hansewissenschaftkolleg (HWK) of Advanced Studies in Delmenhorst, 20-21July 2017. This was the second edition of this symposium which aims to bring together researchers interested in spatial reasoning, problem solving and creativity.
We study the problem of jointly reasoning about language and vision through a navigation and spatial reasoning task. We introduce the Touchdown task and dataset, where an agent must first follow navigation instructions in a real-life visual urban environment, and then identify a location described in natural language to find a hidden object at the goal position. The data contains 9,326 examples of English instructions and spatial descriptions paired with demonstrations. Empirical analysis shows the data presents an open challenge to existing methods, and qualitative linguistic analysis shows that the data displays richer use of spatial reasoning compared to related resources.
Many modern machine learning approaches require vast amounts of training data to learn new concepts; conversely, human learning often requires few examples--sometimes only one--from which the learner can abstract structural concepts. We present a novel approach to introducing new spatial structures to an AI agent, combining deep learning over qualitative spatial relations with various heuristic search algorithms. The agent extracts spatial relations from a sparse set of noisy examples of block-based structures, and trains convolutional and sequential models of those relation sets. To create novel examples of similar structures, the agent begins placing blocks on a virtual table, uses a CNN to predict the most similar complete example structure after each placement, an LSTM to predict the most likely set of remaining moves needed to complete it, and recommends one using heuristic search. We verify that the agent learned the concept by observing its virtual block-building activities, wherein it ranks each potential subsequent action toward building its learned concept. We empirically assess this approach with human participants' ratings of the block structures. Initial results and qualitative evaluations of structures generated by the trained agent show where it has generalized concepts from the training data, which heuristics perform best within the search space, and how we might improve learning and execution.
Natural language processing has made significant inroads into learning the semantics of words through distributional approaches, however representations learnt via these methods fail to capture certain kinds of information implicit in the real world. In particular, spatial relations are encoded in a way that is inconsistent with human spatial reasoning and lacking invariance to viewpoint changes. We present a system capable of capturing the semantics of spatial relations such as behind, left of, etc from natural language. Our key contributions are a novel multi-modal objective based on generating images of scenes from their textual descriptions, and a new dataset on which to train it. We demonstrate that internal representations are robust to meaning preserving transformations of descriptions (paraphrase invariance), while viewpoint invariance is an emergent property of the system.
We introduce cilantro, an open-source C++ library for geometric and general-purpose point cloud data processing. The library provides functionality that covers low-level point cloud operations, spatial reasoning, various methods for point cloud segmentation and generic data clustering, flexible algorithms for robust or local geometric alignment, model fitting, as well as powerful visualization tools. To accommodate all kinds of workflows, cilantro is almost fully templated, and most of its generic algorithms operate in arbitrary data dimension. At the same time, the library is easy to use and highly expressive, promoting a clean and concise coding style. cilantro is highly optimized, has a minimal set of external dependencies, and supports rapid development of performant point cloud processing software in a wide variety of contexts.
Nearly all spatial reasoning problems involve uncertainty of one sort or another. Uncertainty arises due to the inaccuracies of sensors used in measuring distances and angles. We refer to this as directional uncertainty. Uncertainty also arises in combining spatial information when one location is mistakenly identified with another. We refer to this as recognition uncertainty. Most problems in constructing spatial representations (maps) for the purpose of navigation involve both directional and recognition uncertainty. In this paper, we show that a particular class of spatial reasoning problems involving the construction of representations of large-scale space can be solved efficiently even in the presence of directional and recognition uncertainty. We pay particular attention to the problems that arise due to recognition uncertainty.
The study of topological information of spatial objects has for a long time been a focus of research in disciplines like computational geometry, spatial reasoning, cognitive science, and robotics. While the majority of these researches emphasised the topological relations between spatial objects, this work studies the internal topological structure of bounded plane regions, which could consist of multiple pieces and/or have holes and islands to any finite level. The insufficiency of simple regions (regions homeomorphic to closed disks) to cope with the variety and complexity of spatial entities and phenomena has been widely acknowledged. Another significant drawback of simple regions is that they are not closed under set operations union, intersection, and difference. This paper considers bounded semi-algebraic regions, which are closed under set operations and can closely approximate most plane regions arising in practice.
The Region Connection Calculus (RCC) is a well-known calculus for representing part-whole and topological relations. It plays an important role in qualitative spatial reasoning, geographical information science, and ontology. The computational complexity of reasoning with RCC5 and RCC8 (two fragments of RCC) as well as other qualitative spatial/temporal calculi has been investigated in depth in the literature. Most of these works focus on the consistency of qualitative constraint networks. In this paper, we consider the important problem of redundant qualitative constraints. For a set $\Gamma$ of qualitative constraints, we say a constraint $(x R y)$ in $\Gamma$ is redundant if it is entailed by the rest of $\Gamma$. A prime subnetwork of $\Gamma$ is a subset of $\Gamma$ which contains no redundant constraints and has the same solution set as $\Gamma$. It is natural to ask how to compute such a prime subnetwork, and when it is unique. In this paper, we show that this problem is in general intractable, but becomes tractable if $\Gamma$ is over a tractable subalgebra $\mathcal{S}$ of a qualitative calculus. Furthermore, if $\mathcal{S}$ is a subalgebra of RCC5 or RCC8 in which weak composition distributes over nonempty intersections, then $\Gamma$ has a unique prime subnetwork, which can be obtained in cubic time by removing all redundant constraints simultaneously from $\Gamma$. As a byproduct, we show that any path-consistent network over such a distributive subalgebra is weakly globally consistent and minimal. A thorough empirical analysis of the prime subnetwork upon real geographical data sets demonstrates the approach is able to identify significantly more redundant constraints than previously proposed algorithms, especially in constraint networks with larger proportions of partial overlap relations.
Declarative spatial reasoning denotes the ability to (declaratively) specify and solve real-world problems related to geometric and qualitative spatial representation and reasoning within standard knowledge representation and reasoning (KR) based methods (e.g., logic programming and derivatives). One approach for encoding the semantics of spatial relations within a declarative programming framework is by systems of polynomial constraints. However, solving such constraints is computationally intractable in general (i.e. the theory of real-closed fields). We present a new algorithm, implemented within the declarative spatial reasoning system CLP(QS), that drastically improves the performance of deciding the consistency of spatial constraint graphs over conventional polynomial encodings. We develop pruning strategies founded on spatial symmetries that form equivalence classes (based on affine transformations) at the qualitative spatial level. Moreover, pruning strategies are themselves formalised as knowledge about the properties of space and spatial symmetries. We evaluate our algorithm using a range of benchmarks in the class of contact problems, and proofs in mereology and geometry. The empirical results show that CLP(QS) with knowledge-based spatial pruning outperforms conventional polynomial encodings by orders of magnitude, and can thus be applied to problems that are otherwise unsolvable in practice.
The systematic modelling of \emph{dynamic spatial systems} [9] is a key requirement in a wide range of application areas such as comonsense cognitive robotics, computer-aided architecture design, dynamic geographic information systems. We present ASPMT(QS), a novel approach and fully-implemented prototype for non-monotonic spatial reasoning ---a crucial requirement within dynamic spatial systems-- based on Answer Set Programming Modulo Theories (ASPMT). ASPMT(QS) consists of a (qualitative) spatial representation module (QS) and a method for turning tight ASPMT instances into Sat Modulo Theories (SMT) instances in order to compute stable models by means of SMT solvers. We formalise and implement concepts of default spatial reasoning and spatial frame axioms using choice formulas. Spatial reasoning is performed by encoding spatial relations as systems of polynomial constraints, and solving via SMT with the theory of real nonlinear arithmetic. We empirically evaluate ASPMT(QS) in comparison with other prominent contemporary spatial reasoning systems. Our results show that ASPMT(QS) is the only existing system that is capable of reasoning about indirect spatial effects (i.e. addressing the ramification problem), and integrating geometric and qualitative spatial information within a non-monotonic spatial reasoning context.