Recognition of human manipulation actions in real-time is essential for safe and effective human-robot interaction and collaboration. The challenge lies in developing a model that is both lightweight enough for real-time execution and capable of generalization. While some existing methods in the literature can run in real-time, they struggle with temporal scalability, i.e., they fail to adapt to long-duration manipulations effectively. To address this, leveraging the generalizable scene graph representations, we propose a new Factorized Graph Sequence Encoder network that not only runs in real-time but also scales effectively in the temporal dimension, thanks to its factorized encoder architecture. Additionally, we introduce Hand Pooling operation, a simple pooling operation for more focused extraction of the graph-level embeddings. Our model outperforms the previous state-of-the-art real-time approach, achieving a 14.3% and 5.6% improvement in F1-macro score on the KIT Bimanual Action (Bimacs) Dataset and Collaborative Action (CoAx) Dataset, respectively. Moreover, we conduct an extensive ablation study to validate our network design choices. Finally, we compare our model with its architecturally similar RGB-based model on the Bimacs dataset and show the limitations of this model in contrast to ours on such an object-centric manipulation dataset. Our code and trained models are available at https://github.com/eneserdo/FGSE.

In this work, we propose a novel, fast, and memory-efficient unsupervised statistical method, combined with an unsupervised deep learning (DL) model, for de-snowing 3D LiDAR point clouds in a fully unsupervised fashion. The results obtained on the real-scanned Winter Adverse Driving dataSet (WADS) show that our DL model achieves a 6.3% improvement in mIoU over the current state-of-the-art unsupervised DL methods and performs comparable to supervised counterparts, substantially narrowing the performance gap between supervised and unsupervised approaches. In addition to that, our model also outperforms its closest competitor by 12.8% mIoU when tested on our Canadian Adverse Driving Conditions (CADC) dataset annotations. Additionally, our de-snowing algorithm enhances downstream semantic segmentation and object detection tasks without even requiring any modifications to the base segmentation and detection models. The source code, trained models, and the online supplementary information are available at the following URL: https://sporsho.github.io/3DUnOutDet.

Although LiDAR sensors are crucial for autonomous systems due to providing precise depth information, they struggle with capturing fine object details, especially at a distance, due to sparse and non-uniform data. Recent advances introduced pseudo-LiDAR, i.e., synthetic dense point clouds, using additional modalities such as cameras to enhance 3D object detection. We present a novel LiDAR-only framework that augments raw scans with denser pseudo point clouds by solely relying on LiDAR sensors and scene semantics, omitting the need for cameras. Our framework first utilizes a segmentation model to extract scene semantics from raw point clouds, and then employs a multi-modal domain translator to generate synthetic image segments and depth cues without real cameras. This yields a dense pseudo point cloud enriched with semantic information. We also introduce a new semantically guided projection method, which enhances detection performance by retaining only relevant pseudo points. We applied our framework to different advanced 3D object detection methods and reported up to 2.9% performance upgrade. We also obtained comparable results on the KITTI 3D object detection test set, in contrast to other state-of-the-art LiDAR-only detectors.

Adverse weather conditions such as snow, rain, and fog are natural phenomena that can impair the performance of the perception algorithms in autonomous vehicles. Although LiDARs provide accurate and reliable scans of the surroundings, its output can be substantially degraded by precipitation (e.g., snow particles) leading to an undesired effect on the downstream perception tasks. Several studies have been performed to battle this undesired effect by filtering out precipitation outliers, however, these works have large memory consumption and long execution times which are not desired for onboard applications. To that end, we introduce a novel outlier detector for 3D LiDAR point clouds captured under adverse weather conditions. Our proposed detector 3D-OutDet is based on a novel convolution operation that processes nearest neighbors only, allowing the model to capture the most relevant points. This reduces the number of layers, resulting in a model with a low memory footprint and fast execution time, while producing a competitive performance compared to state-of-the-art models. We conduct extensive experiments on three different datasets (WADS, SnowyKITTI, and SemanticSpray) and show that with a sacrifice of 0.16% mIOU performance, our model reduces the memory consumption by 99.92%, number of operations by 96.87%, and execution time by 82.84% per point cloud on the real-scanned WADS dataset. Our experimental evaluations also showed that the mIOU performance of the downstream semantic segmentation task on WADS can be improved up to 5.08% after applying our proposed outlier detector. We release our source code, supplementary material and videos in https://sporsho.github.io/3DOutDet. Upon clicking the link you will have to option to go to source code, see supplementary information and view videos generated with our 3D-OutDet.

In this work, we present a new single-stage method for subject agnostic face swapping and identity transfer, named FaceDancer. We have two major contributions: Adaptive Feature Fusion Attention (AFFA) and Interpreted Feature Similarity Regularization (IFSR). The AFFA module is embedded in the decoder and adaptively learns to fuse attribute features and features conditioned on identity information without requiring any additional facial segmentation process. In IFSR, we leverage the intermediate features in an identity encoder to preserve important attributes such as head pose, facial expression, lighting, and occlusion in the target face, while still transferring the identity of the source face with high fidelity. We conduct extensive quantitative and qualitative experiments on various datasets and show that the proposed FaceDancer outperforms other state-of-the-art networks in terms of identity transfer, while having significantly better pose preservation than most of the previous methods. Code available at https://github.com/felixrosberg/FaceDance.

The compression of deep learning models is of fundamental importance in deploying such models to edge devices. The selection of compression parameters can be automated to meet changes in the hardware platform and application. This article introduces a Multi-Objective Hardware-Aware Quantization (MOHAQ) method, which considers hardware performance and inference error as objectives for mixed-precision quantization. The proposed method feasibly evaluates candidate solutions in a large search space by relying on two steps. First, post-training quantization is applied for fast solution evaluation (inference-only search). Second, we propose the ”beacon-based search” to retrain selected solutions only and use them as beacons to estimate the effect of retraining on other solutions. We use speech recognition models on TIMIT dataset. Experimental evaluations show that Simple Recurrent Unit (SRU)-based models can be compressed up to 8x by post-training quantization without any significant error increase. On SiLago, we found solutions that achieve 97% and 86% of the maximum possible speedup and energy saving, with a minor increase in error on an SRU-based model. On Bitfusion, the beacon-based search reduced the error gain of the inference-only search on SRU-based models and Light Gated Recurrent Unit (LiGRU)-based model by up to 4.9 and 3.9 percentage points, respectively.

We need robots more aware of the unintended outcomes of their actions for ensuring safety. This can be achieved by an onboard failure detection system to monitor and detect such cases. Onboard failure detection is challenging with a limited set of onboard sensor setup due to the limitations of sensing capabilities of each sensor. To alleviate these challenges, we propose FINO-Net, a novel multimodal sensor fusion based deep neural network to detect and identify manipulation failures. We also introduce FAILURE, a multimodal dataset, containing 229 real-world manipulation data recorded with a Baxter robot. Our network combines RGB, depth and audio readings to effectively detect failures. Results indicate that fusing RGB with depth and audio modalities significantly improves the performance. FINO-Net achieves %98.60 detection accuracy on our novel dataset. Code and data are publicly available at https://github.com/ardai/fino-net.

In this paper, we introduce a deep encoder-decoder network, named SalsaNet, for efficient semantic segmentation of 3D LiDAR point clouds. SalsaNet segments the road, i.e. drivable free-space, and vehicles in the scene by employing the Bird-Eye-View (BEV) image projection of the point cloud. To overcome the lack of annotated point cloud data, in particular for the road segments, we introduce an auto-labeling process which transfers automatically generated labels from the camera to LiDAR. We also explore the role of image-like projection of LiDAR data in semantic segmentation by comparing BEV with spherical-front-view projection and show that SalsaNet is projection-agnostic. We perform quantitative and qualitative evaluations on the KITTI dataset, which demonstrate that the proposed SalsaNet outperforms other state-of-the-art semantic segmentation networks in terms of accuracy and computation time. Our code and data are publicly available at https://gitlab.com/aksoyeren/salsanet.git.

Socially assistive robots are increasingly being designed to interact with humans in various therapeutical scenarios. We believe that one useful scenario is providing exercise coaching for Persons with Dementia (PWD), which involves unique challenges related to memory and communication. We present a design for a robot that can seek to help a PWD to conduct exercises by recognizing their behaviors and providing appropriate feedback, in an online, multimodal, and engaging way. Additionally, following a mid-fidelity prototyping approach, we report on some observations from an exploratory user study using a Baxter robot; although limited by the sample size and our simplified approach, the results suggested the usefulness of the general scenario, and that the degree to which a robot provides feedback-occasional or continuous- could moderate impressions of attentiveness or fun. Some possibilities for future improvement are outlined, touching on richer recognition and behavior generation strategies based on deep learning and haptic feedback, toward informing next designs.

Neural networks have been notorious for being computational expensive. This is mainly because neural networks are often over-parametrized and most likely have redundant nodes or layers as they are getting deeper and wider. Their demand for hardware resources prohibits their extensive use in embedded devices and puts restrictions on tasks like real time image classification or object detection. In this work, we propose a network agnostic model compression method infused with a novel dynamical clustering approach to reduce the computational cost and memory footprint of deep neural networks. We evaluated our new compression method on five different state-of-the-art image classification and object detection networks. In classification networks, we pruned about 95% of network parameters. In advanced detection networks such as YOLOv3, our proposed compression method managed to reduce the model parameters up to 59.70% which yielded 110× less memory without sacrificing much in accuracy.

We present a three-level cognitive system in a learning by demonstration context. The system allows for learning and transfer on the sensorimotor level as well as the planning level. The fundamentally different data structures associated with these two levels are connected by an efficient mid-level representation based on so-called “semantic event chains.” We describe details of the representations and quantify the effect of the associated learning procedures for each level under different amounts of noise. Moreover, we demonstrate the performance of the overall system by three demonstrations that have been performed at a project review. The described system has a technical readiness level (TRL) of 4, which in an ongoing follow-up project will be raised to TRL 6.

Recognition of human manipulation actions together with the analysis and execution by a robot is an important issue. Also, perception of spatial relationships between objects is central to understanding the meaning of manipulation actions. Here we would like to merge these two notions and analyze manipulation actions using symbolic spatial relations between objects in the scene. Specifically, we define procedures for extraction of symbolic human-readable relations based on Axis Aligned Bounding Box object models and use sequences of those relations for action recognition from image sequences. Our framework is inspired by the so called Semantic Event Chain framework, which analyzes touching and un-touching events of different objects during the manipulation. However, our framework uses fourteen spatial relations instead of two. We show that our relational framework is able to differentiate between more manipulation actions than the original Semantic Event Chains. We quantitatively evaluate the method on the MANIAC dataset containing 120 videos of eight different manipulation actions and obtain 97% classification accuracy which is 12 % more as compared to the original Semantic Event Chains.

We propose a formalism for the hierarchical representation of affordances. Starting with a perceived model of the environment consisting of geometric primitives like planes or cylinders, we define a hierarchical system for affordance extraction whose foundation are elementary power grasp affordances. Higher-level affordances, e.g. bimanual affordances, result from combining lower-level affordances with additional properties concerning the underlying geometric primitives of the scene. We model affordances as continuous certainty functions taking into account properties of the environmental elements and the perceiving robot's embodiment. The developed formalism is regarded as the basis for the description of whole-body affordances, i.e. affordances associated with whole-body actions. The proposed formalism was implemented and experimentally evaluated in multiple scenarios based on RGB-D camera data. The feasibility of the approach is demonstrated on a real robotic platform.

In this paper we present a formal computational framework for modeling manipulation actions. The introduced formalism leads to semantics of manipulation action and has applications to both observing and understanding human manipulation actions as well as executing them with a robotic mechanism (e.g. a humanoid robot). It is based on a Combinatory Categorial Grammar. The goal of the introduced framework is to: (1) represent manipulation actions with both syntax and semantic parts, where the semantic part employs $\lambda$-calculus; (2) enable a probabilistic semantic parsing schema to learn the $\lambda$-calculus representation of manipulation action from an annotated action corpus of videos; (3) use (1) and (2) to develop a system that visually observes manipulation actions and understands their meaning while it can reason beyond observations using propositional logic and axiom schemata. The experiments conducted on a public available large manipulation action dataset validate the theoretical framework and our implementation.

In this work we address the problem of finding replacements of missing objects that are needed for the execution of human-like manipulation tasks. This is a usual problem that is easily solved by humans provided their natural knowledge to find object substitutions: using a knife as a screwdriver or a book as a cutting board. On the other hand, in robotic applications, objects required in the task should be included in advance in the problem definition. If any of these objects is missing from the scenario, the conventional approach is to manually redefine the problem according to the available objects in the scene. In this work we propose an automatic way of finding object substitutions for the execution of manipulation tasks. The approach uses a logic-based planner to generate a plan from a prototypical problem definition and searches for replacements in the scene when some of the objects involved in the plan are missing. This is done by means of a repository of objects and attributes with roles, which is used to identify the affordances of the unknown objects in the scene. Planning actions are grounded using a novel approach that encodes the semantic structure of manipulation actions. The system was evaluated in a KUKA arm platform for the task of preparing a salad with successful results.

Human manipulation activity recognition is an important yet challenging task in robot imitation. In this paper, we introduce, for the first time, a novel method for semantic decomposition and recognition of continuous human manipulation activities by using on-line learned individual manipulation models. Solely based on the spatiotemporal interactions between objects and hands in the scene, the proposed framework can parse not only sequential and concurrent (overlapping) manipulation streams but also basic primitive elements of each detected manipulation. Without requiring any prior object knowledge, the framework can furthermore extract object-like scene entities that are performing the same role in the detected manipulations. The framework was evaluated on our new egocentric activity dataset which contains 120 different samples of 8 single atomic manipulations (e.g. Cutting and Stirring) and 20 long and complex activity demonstrations such as “making a sandwich” and “preparing a breakfast”. We finally show that parsed manipulation actions can be imitated by robots even in various scene contexts with novel objects.

We describe a system allowing a robot to learn goal-directed manipulation sequences such as steps of an assembly task. Learning is based on a free mix of exploration and instruction by an external teacher, and may be active in the sense that the system tests actions to maximize learning progress and asks the teacher if needed. The main component is a symbolic planning engine that operates on learned rules, defined by actions and their pre- and postconditions. Learned by model-based reinforcement learning, rules are immediately available for planning. Thus, there are no distinct learning and application phases. We show how dynamic plans, replanned after every action if necessary, can be used for automatic execution of manipulation sequences, for monitoring of observed manipulation sequences, or a mix of the two, all while extending and refining the rule base on the fly. Quantitative results indicate fast convergence using few training examples, and highly effective teacher intervention at early stages of learning.

The goal of this study is to provide an architecture for a generic definition of robot manipulation actions. We emphasize that the representation of actions presented here is “procedural”. Thus, we will define the structural elements of our action representations as execution protocols. To achieve this, manipulations are defined using three levels. The toplevel defines objects, their relations and the actions in an abstract and symbolic way. A mid-level sequencer, with which the action primitives are chained, is used to structure the actual action execution, which is performed via the bottom level. This (lowest) level collects data from sensors and communicates with the control system of the robot. This method enables robot manipulators to execute the same action in different situations i.e. on different objects with different positions and orientations. In addition, two methods of detecting action failure are provided which are necessary to handle faults in system. To demonstrate the effectiveness of the proposed framework, several different actions are performed on our robotic setup and results are shown. This way we are creating a library of human-like robot actions, which can be used by higher-level task planners to execute more complex tasks.

Robust visual tracking is an essential precursor to understanding and replicating human actions in robotic systems. In order to accurately evaluate the semantic meaning of a sequence of video frames, or to replicate an action contained therein, one must be able to coherently track and segment all observed agents and objects. This work proposes a novel online point cloud based algorithm which simultaneously tracks 6DoF pose and determines spatial extent of all entities in indoor scenarios. This is accomplished using a persistent supervoxel world-model which is updated, rather than replaced, as new frames of data arrive. Maintenance of a world model enables general object permanence, permitting successful tracking through full occlusions. Object models are tracked using a bank of independent adaptive particle filters which use a supervoxel observation model to give rough estimates of object state. These are united using a novel multi-model RANSAC-like approach, which seeks to minimize a global energy function associating world-model supervoxels to predicted states. We present results on a standard robotic assembly benchmark for two application scenarios - human trajectory imitation and semantic action understanding - demonstrating the usefulness of the tracking in intelligent robotic systems.

We present an architecture for real-time, online vision systems which enables development and use of complex vision pipelines integrating any number of algorithms. Individual algorithms are implemented using modular plugins, allowing integration of independently developed algorithms and rapid testing of new vision pipeline configurations. The architecture exploits the parallelization of graphics processing units (GPUs) and multi-core systems to speed processing and achieve real-time performance. Additionally, the use of a global memory management system for frame buffering permits complex algorithmic flow (e.g. feedback loops) in online processing setups, while maintaining the benefits of threaded asynchronous operation of separate algorithms. To demonstrate the system, a typical real-time system setup is described which incorporates plugins for video and depth acquisition, GPU-based segmentation and optical flow, semantic graph generation, and online visualization of output. Performance numbers are shown which demonstrate the insignificant overhead cost of the architecture as well as speed-up over strictly CPU and single threaded implementations.

In this work we introduce a novel approach for detecting spatiotemporal object-action relations, leading to both, action recognition and object categorization. Semantic scene graphs are extracted from image sequences and used to find the characteristic main graphs of the action sequence via an exact graph-matching technique, thus providing an event table of the action scene, which allows extracting object-action relations. The method is applied to several artificial and real action scenes containing limited context. The central novelty of this approach is that it is model free and needs a priori representation neither for objects nor actions. Essentially actions are recognized without requiring prior object knowledge and objects are categorized solely based on their exhibited role within an action sequence. Thus, this approach is grounded in the affordance principle, which has recently attracted much attention in robotics and provides a way forward for trial and error learning of object-action relations through repeated experimentation. It may therefore be useful for recognition and categorization tasks for example in imitation learning in developmental and cognitive robotics.