Physical construction---the ability to compose objects, subject to physical dynamics, to serve some function---is fundamental to human intelligence. We introduce a suite of challenging physical construction tasks inspired by how children play with blocks, such as matching a target configuration, stacking blocks to connect objects together, and creating shelter-like structures over target objects. We examine how a range of deep reinforcement learning agents fare on these challenges, and introduce several new approaches which provide superior performance. Our results show that agents which use structured representations (e.g., objects and scene graphs) and structured policies (e.g., object-centric actions) outperform those which use less structured representations, and generalize better beyond their training when asked to reason about larger scenes. Model-based agents which use Monte-Carlo Tree Search also outperform strictly model-free agents in our most challenging construction problems. We conclude that approaches which combine structured representations and reasoning with powerful learning are a key path toward agents that possess rich intuitive physics, scene understanding, and planning.
An important goal of research in Deep Reinforcement Learning in mobile robotics is to train agents capable of solving complex tasks, which require a high level of scene understanding and reasoning from an egocentric perspective. When trained from simulations, optimal environments should satisfy a currently unobtainable combination of high-fidelity photographic observations, massive amounts of different environment configurations and fast simulation speeds. In this paper we argue that research on training agents capable of complex reasoning can be simplified by decoupling from the requirement of high fidelity photographic observations. We present a suite of tasks requiring complex reasoning and exploration in continuous, partially observable 3D environments. The objective is to provide challenging scenarios and a robust baseline agent architecture that can be trained on mid-range consumer hardware in under 24h. Our scenarios combine two key advantages: (i) they are based on a simple but highly efficient 3D environment (ViZDoom) which allows high speed simulation (12000fps); (ii) the scenarios provide the user with a range of difficulty settings, in order to identify the limitations of current state of the art algorithms and network architectures. We aim to increase accessibility to the field of Deep-RL by providing baselines for challenging scenarios where new ideas can be iterated on quickly. We argue that the community should be able to address challenging problems in reasoning of mobile agents without the need for a large compute infrastructure.
Semantic segmentation generates comprehensive understanding of scenes through densely predicting the category for each pixel. High-level features from Deep Convolutional Neural Networks already demonstrate their effectiveness in semantic segmentation tasks, however the coarse resolution of high-level features often leads to inferior results for small/thin objects where detailed information is important. It is natural to consider importing low level features to compensate for the lost detailed information in high-level features.Unfortunately, simply combining multi-level features suffers from the semantic gap among them. In this paper, we propose a new architecture, named Gated Fully Fusion (GFF), to selectively fuse features from multiple levels using gates in a fully connected way. Specifically, features at each level are enhanced by higher-level features with stronger semantics and lower-level features with more details, and gates are used to control the propagation of useful information which significantly reduces the noises during fusion. We achieve the state of the art results on four challenging scene parsing datasets including Cityscapes, Pascal Context, COCO-stuff and ADE20K.
Semantic scene understanding is important for various applications. In particular, self-driving cars need a fine-grained understanding of the surfaces and objects in their vicinity. Light detection and ranging (LiDAR) provides precise geometric information about the environment and is thus a part of the sensor suites of almost all self-driving cars. Despite the relevance of semantic scene understanding for this application, there is a lack of a large dataset for this task which is based on an automotive LiDAR. In this paper, we introduce a large dataset to propel research on laser-based semantic segmentation. We annotated all sequences of the KITTI Vision Odometry Benchmark and provide dense point-wise annotations for the complete $360^{o}$ field-of-view of the employed automotive LiDAR. We propose three benchmark tasks based on this dataset: (i) semantic segmentation of point clouds using a single scan, (ii) semantic segmentation using multiple past scans, and (iii) semantic scene completion, which requires to anticipate the semantic scene in the future. We provide baseline experiments and show that there is a need for more sophisticated models to efficiently tackle these tasks. Our dataset opens the door for the development of more advanced methods, but also provides plentiful data to investigate new research directions.
Deep learning techniques have become the to-go models for most vision-related tasks on 2D images. However, their power has not been fully realised on several tasks in 3D space, e.g., 3D scene understanding. In this work, we jointly address the problems of semantic and instance segmentation of 3D point clouds. Specifically, we develop a multi-task pointwise network that simultaneously performs two tasks: predicting the semantic classes of 3D points and embedding the points into high-dimensional vectors so that points of the same object instance are represented by similar embeddings. We then propose a multi-value conditional random field model to incorporate the semantic and instance labels and formulate the problem of semantic and instance segmentation as jointly optimising labels in the field model. The proposed method is thoroughly evaluated and compared with existing methods on different indoor scene datasets including S3DIS and SceneNN. Experimental results showed the robustness of the proposed joint semantic-instance segmentation scheme over its single components. Our method also achieved state-of-the-art performance on semantic segmentation.
Scene understanding of high resolution aerial images is of great importance for the task of automated monitoring in various remote sensing applications. Due to the large within-class and small between-class variance in pixel values of objects of interest, this remains a challenging task. In recent years, deep convolutional neural networks have started being used in remote sensing applications and demonstrate state of the art performance for pixel level classification of objects. \textcolor{black}{Here we propose a reliable framework for performant results for the task of semantic segmentation of monotemporal very high resolution aerial images. Our framework consists of a novel deep learning architecture, ResUNet-a, and a novel loss function based on the Dice loss. ResUNet-a uses a UNet encoder/decoder backbone, in combination with residual connections, atrous convolutions, pyramid scene parsing pooling and multi-tasking inference. ResUNet-a infers sequentially the boundary of the objects, the distance transform of the segmentation mask, the segmentation mask and a colored reconstruction of the input. Each of the tasks is conditioned on the inference of the previous ones, thus establishing a conditioned relationship between the various tasks, as this is described through the architecture's computation graph. We analyse the performance of several flavours of the Generalized Dice loss for semantic segmentation, and we introduce a novel variant loss function for semantic segmentation of objects that has excellent convergence properties and behaves well even under the presence of highly imbalanced classes.} The performance of our modeling framework is evaluated on the ISPRS 2D Potsdam dataset. Results show state-of-the-art performance with an average F1 score of 92.9\% over all classes for our best model.
Occupancy grid mapping is an important component in road scene understanding for autonomous driving. It encapsulates information of the drivable area, road obstacles and enables safe autonomous driving. Radars are an emerging sensor in autonomous vehicle vision, becoming more widely used due to their long range sensing, low cost, and robustness to severe weather conditions. Despite recent advances in deep learning technology, occupancy grid mapping from radar data is still mostly done using classical filtering approaches.In this work, we propose learning the inverse sensor model used for occupancy grid mapping from clustered radar data. This is done in a data driven approach that leverages computer vision techniques. This task is very challenging due to data sparsity and noise characteristics of the radar sensor. The problem is formulated as a semantic segmentation task and we show how it can be learned using lidar data for generating ground truth. We show both qualitatively and quantitatively that our learned occupancy net outperforms classic methods by a large margin using the recently released NuScenes real-world driving data.
Generating images via the generative adversarial network (GAN) has attracted much attention recently. However, most of the existing GAN-based methods can only produce low-resolution images of limited quality. Directly generating high-resolution images using GANs is nontrivial, and often produces problematic images with incomplete objects. To address this issue, we develop a novel GAN called Auto-Embedding Generative Adversarial Network (AEGAN), which simultaneously encodes the global structure features and captures the fine-grained details. In our network, we use an autoencoder to learn the intrinsic high-level structure of real images and design a novel denoiser network to provide photo-realistic details for the generated images. In the experiments, we are able to produce 512x512 images of promising quality directly from the input noise. The resultant images exhibit better perceptual photo-realism, i.e., with sharper structure and richer details, than other baselines on several datasets, including Oxford-102 Flowers, Caltech-UCSD Birds (CUB), High-Quality Large-scale CelebFaces Attributes (CelebA-HQ), Large-scale Scene Understanding (LSUN) and ImageNet.
Robust geometric and semantic scene understanding is ever more important in many real-world applications such as autonomous driving and robotic navigation. In this paper, we propose a multi-task learning-based approach capable of jointly performing geometric and semantic scene understanding, namely depth prediction (monocular depth estimation and depth completion) and semantic scene segmentation. Within a single temporally constrained recurrent network, our approach uniquely takes advantage of a complex series of skip connections, adversarial training and the temporal constraint of sequential frame recurrence to produce consistent depth and semantic class labels simultaneously. Extensive experimental evaluation demonstrates the efficacy of our approach compared to other contemporary state-of-the-art techniques.
Monocular 3D scene understanding tasks, such as object size estimation, heading angle estimation and 3D localization, is challenging. Successful modern day methods for 3D scene understanding require the use of a 3D sensor. On the other hand, single image based methods have significantly worse performance. In this work, we aim at bridging the performance gap between 3D sensing and 2D sensing for 3D object detection by enhancing LiDAR-based algorithms to work with single image input. Specifically, we perform monocular depth estimation and lift the input image to a point cloud representation, which we call pseudo-LiDAR point cloud. Then we can train a LiDAR-based 3D detection network with our pseudo-LiDAR end-to-end. Following the pipeline of two-stage 3D detection algorithms, we detect 2D object proposals in the input image and extract a point cloud frustum from the pseudo-LiDAR for each proposal. Then an oriented 3D bounding box is detected for each frustum. To handle the large amount of noise in the pseudo-LiDAR, we propose two innovations: (1) use a 2D-3D bounding box consistency constraint, adjusting the predicted 3D bounding box to have a high overlap with its corresponding 2D proposal after projecting onto the image; (2) use the instance mask instead of the bounding box as the representation of 2D proposals, in order to reduce the number of points not belonging to the object in the point cloud frustum. Through our evaluation on the KITTI benchmark, we achieve the top-ranked performance on both bird's eye view and 3D object detection among all monocular methods, effectively quadrupling the performance over previous state-of-the-art. Our code is available at https://github.com/xinshuoweng/Mono3D_PLiDAR.
Over the past few years, deep learning techniques have achieved tremendous success in many visual understanding tasks such as object detection, image segmentation, and caption generation. Despite this thriving in computer vision and natural language processing, deep learning has not yet shown significant impact in robotics. Due to the gap between theory and application, there are many challenges when applying the results of deep learning to the real robotic systems. In this study, our long-term goal is to bridge the gap between computer vision and robotics by developing visual methods that can be used in real robots. In particular, this work tackles two fundamental visual problems for autonomous robotic manipulation: affordance detection and fine-grained action understanding. Theoretically, we propose different deep architectures to further improves the state of the art in each problem. Empirically, we show that the outcomes of our proposed methods can be applied in real robots and allow them to perform useful manipulation tasks.
Mobile robots that manipulate their environments require high-accuracy scene understanding at close range. Typically this understanding is achieved with RGBD cameras, but the evaluation process for selecting an appropriate RGBD camera for the application is minimally quantitative. Limited manufacturer-published metrics do not translate to observed quality in real-world cluttered environments, since quality is application-specific. To bridge the gap, we present a method for quantitatively measuring depth quality using a set of extendable 3D printed fixtures that approximate real-world conditions. By framing depth quality as point cloud density and root mean square error (RMSE) from a known geometry, we present a method that is extendable by other system integrators for custom environments. We show a comparison of 3 cameras and present a case study for camera selection, provide reference meshes and analysis code, and discuss further extensions.
Recent development in autonomous driving involves high-level computer vision and detailed road scene understanding. Today, most autonomous vehicles are using mediated perception approach for path planning and control, which highly rely on high-definition 3D maps and real time sensors. Recent research efforts aim to substitute the massive HD maps with coarse road attributes. In this paper, we follow the direct perception based method to train a deep neural network for affordance learning in autonomous driving. Our goal in this work is to develop the affordance learning model based on freely available Google Street View panoramas and Open Street Map road vector attributes. Driving scene understanding can be achieved by learning affordances from the images captured by car-mounted cameras. Such scene understanding by learning affordances may be useful for corroborating base maps such as HD maps so that the required data storage space is minimized and available for processing in real time. We compare capability in road attribute identification between human volunteers and our model by experimental evaluation. Our results indicate that this method could act as a cheaper way for training data collection in autonomous driving. The cross validation results also indicate the effectiveness of our model.
The rapid development of autonomous driving in recent years presents lots of challenges for scene understanding. As an essential step towards scene understanding, semantic segmentation thus received lots of attention in past few years. Although deep learning based state-of-the-arts have achieved great success in improving the segmentation accuracy, most of them suffer from an inefficiency problem and can hardly applied to practical applications. In this paper, we systematically analyze the computation cost of Convolutional Neural Network(CNN) and found that the inefficiency of CNN is mainly caused by its wide structure rather than the deep structure. In addition, the success of pruning based model compression methods proved that there are many redundant channels in CNN. Thus, we designed a very narrow while deep backbone network to improve the efficiency of semantic segmentation. By casting our network to FCN32 segmentation architecture, the basic structure of most segmentation methods, we achieved 60.6\% mIoU on Cityscape val dataset with 54 frame per seconds(FPS) on $1024\times2048$ inputs, which already outperforms one of the earliest real time deep learning based segmentation methods: ENet.
6D object pose estimation is an important task that determines the 3D position and 3D rotation of an object in camera-centred coordinates. By utilizing such a task, one can propose promising solutions for various problems related to scene understanding, augmented reality, control and navigation of robotics. Recent developments on visual depth sensors and low-cost availability of depth data significantly facilitate object pose estimation. Using depth information from RGB-D sensors, substantial progress has been made in the last decade by the methods addressing the challenges such as viewpoint variability, occlusion and clutter, and similar looking distractors. Particularly, with the recent advent of convolutional neural networks, RGB-only based solutions have been presented. However, improved results have only been reported for recovering the pose of known instances, i.e., for the instance-level object pose estimation tasks. More recently, state-of-the-art approaches target to solve object pose estimation problem at the level of categories, recovering the 6D pose of unknown instances. To this end, they address the challenges of the category-level tasks such as distribution shift among source and target domains, high intra-class variations, and shape discrepancies between objects.
Indoor scenes exhibit rich hierarchical structure in 3D object layouts. Many tasks in 3D scene understanding can benefit from reasoning jointly about the hierarchical context of a scene, and the identities of objects. We present a variational denoising recursive autoencoder (VDRAE) that generates and iteratively refines a hierarchical representation of 3D object layouts, interleaving bottom-up encoding for context aggregation and top-down decoding for propagation. We train our VDRAE on large-scale 3D scene datasets to predict both instance-level segmentations and a 3D object detections from an over-segmentation of an input point cloud. We show that our VDRAE improves object detection performance on real-world 3D point cloud datasets compared to baselines from prior work.
Robots need to understand their environment to perform their task. If it is possible to pre-program a visual scene analysis process in closed environments, robots operating in an open environment would benefit from the ability to learn it through their interaction with their environment. This ability furthermore opens the way to the acquisition of affordances maps in which the action capabilities of the robot structure its visual scene understanding. We propose an approach to build such affordances maps by relying on an interactive perception approach and an online classification. In the proposed formalization of affordances, actions and effects are related to visual features, not objects, and they can be combined. We have tested the approach on three action primitives and on a real PR2 robot.
3D multi-object detection and tracking are crucial for traffic scene understanding. However, the community pays less attention to these areas due to the lack of a standardized benchmark dataset to advance the field. Moreover, existing datasets (e.g., KITTI) do not provide sufficient data and labels to tackle challenging scenes where highly interactive and occluded traffic participants are present. To address the issues, we present the Honda Research Institute 3D Dataset (H3D), a large-scale full-surround 3D multi-object detection and tracking dataset collected using a 3D LiDAR scanner. H3D comprises of 160 crowded and highly interactive traffic scenes with a total of 1 million labeled instances in 27,721 frames. With unique dataset size, rich annotations, and complex scenes, H3D is gathered to stimulate research on full-surround 3D multi-object detection and tracking. To effectively and efficiently annotate a large-scale 3D point cloud dataset, we propose a labeling methodology to speed up the overall annotation cycle. A standardized benchmark is created to evaluate full-surround 3D multi-object detection and tracking algorithms. 3D object detection and tracking algorithms are trained and tested on H3D. Finally, sources of errors are discussed for the development of future algorithms.
Assigning a label to each pixel in an image, namely semantic segmentation, has been an important task in computer vision, and has applications in autonomous driving, robotic navigation, localization, and scene understanding. Fully convolutional neural networks have proved to be a successful solution for the task over the years but most of the work being done focuses primarily on accuracy. In this paper, we present a computationally efficient approach to semantic segmentation, while achieving a high mean intersection over union (mIOU), 70.33% on Cityscapes challenge. The network proposed is capable of running real-time on mobile devices. In addition, we make our code and model weights publicly available.
Detecting visual relationships, i.e. <Subject, Predicate, Object> triplets, is a challenging Scene Understanding task approached in the past via linguistic priors or spatial information in a single feature branch. We introduce a new deeply supervised two-branch architecture, the Multimodal Attentional Translation Embeddings, where the visual features of each branch are driven by a multimodal attentional mechanism that exploits spatio-linguistic similarities in a low-dimensional space. We present a variety of experiments comparing against all related approaches in the literature, as well as by re-implementing and fine-tuning several of them. Results on the commonly employed VRD dataset [1] show that the proposed method clearly outperforms all others, while we also justify our claims both quantitatively and qualitatively.