Object Pose Research Articles

CPPF++: Uncertainty-Aware Sim2Real Object Pose Estimation by Vote Aggregation.

Object pose estimation constitutes a critical area within the domain of 3D vision. While contemporary state-of-the-art methods that leverage real-world pose annotations have demonstrated commendable performance, the procurement of such real training data incurs substantial costs. This paper focuses on a specific setting wherein only 3D CAD models are utilized as a priori knowledge, devoid of any background or clutter information. We introduce a novel method, CPPF++, designed for sim-to-real category-level pose estimation. This method builds upon the foundational point-pair voting scheme of CPPF, reformulating it through a probabilistic view. To address the challenge posed by vote collision, we propose a novel approach that involves modeling the voting uncertainty by estimating the probabilistic distribution of each point pair within the canonical space. Furthermore, we augment the contextual information provided by each voting unit through the introduction of N-point tuples. To enhance the robustness and accuracy of the model, we incorporate several innovative modules, including noisy pair filtering, online alignment optimization, and a tuple feature ensemble. Alongside these methodological advancements, we introduce a new category-level pose estimation dataset, named DiversePose 300. Empirical evidence demonstrates that our method significantly surpasses previous sim-to-real approaches and achieves comparable or superior performance on novel datasets.

NeuralFeels with neural fields: Visuotactile perception for in-hand manipulation.

To achieve human-level dexterity, robots must infer spatial awareness from multimodal sensing to reason over contact interactions. During in-hand manipulation of novel objects, such spatial awareness involves estimating the object's pose and shape. The status quo for in-hand perception primarily uses vision and is restricted to tracking a priori known objects. Moreover, visual occlusion of objects in hand is imminent during manipulation, preventing current systems from pushing beyond tasks without occlusion. We combined vision and touch sensing on a multifingered hand to estimate an object's pose and shape during in-hand manipulation. Our method, NeuralFeels, encodes object geometry by learning a neural field online and jointly tracks it by optimizing a pose graph problem. We studied multimodal in-hand perception in simulation and the real world, interacting with different objects via a proprioception-driven policy. Our experiments showed final reconstruction F scores of 81% and average pose drifts of 4.7 millimeters, which was further reduced to 2.3 millimeters with known object models. In addition, we observed that, under heavy visual occlusion, we could achieve improvements in tracking up to 94% compared with vision-only methods. Our results demonstrate that touch, at the very least, refines and, at the very best, disambiguates visual estimates during in-hand manipulation. We release our evaluation dataset of 70 experiments, FeelSight, as a step toward benchmarking in this domain. Our neural representation driven by multimodal sensing can serve as a perception backbone toward advancing robot dexterity.

TopoPIS: Topology-constrained pipe instance segmentation via adaptive curvature convolution

Precise and fast pipe instance segmentation is a critical component in industrial automatic assembly, facilitating accurate object detection and pose estimation, optimizing and supervising the assembly process. However, this problem is challenging due to topological errors on fine-scale structures caused by the pipes being complex and slender. To address these challenges, we propose a topology-constrained pipe instance segmentation network (TopoPIS) for complex stacking scene to achieve accurate segmentation with topological correctness. To better extract the features of complex and variable morphological pipes, adaptive curvature convolution is introduced to dynamically adapt to the slender pipe structure and capture critical features. To handle topological errors like broken connections, we propose a novel topological constraint loss function based on persistent homology, which greatly improves the topological continuity of the segmentation. Experimental results on real-world and unseen datasets demonstrate that our TopoPIS outperforms other methods regrading segmentation accuracy and topological continuity.

Cross-modal attention and geometric contextual aggregation network for 6DoF object pose estimation

Cross-modal attention and geometric contextual aggregation network for 6DoF object pose estimation

Mетод визначення зміни просторового положення об'єкта орбітального сервісу з невідомою формою

This paper presents a method for determining the pose of an on-orbit service object (target) with an unknown surface shape based on processing target surface point clouds. It is assumed that a point cloud is obtained using a sensor in the form of a lidar-based system onboard the service spacecraft. The algorithm of the method operates with a simplified description of the cloud; it uses only the coordinates of the cloud points without identifying any surface features. The idea of the method is as follows. From all cloud points obtained at a certain time instant, select points that are close to a certain degree to a given set of planes arranged in space in a certain way. From the point cloud corresponding to the next time instant, select points that are close to the same set of planes, but moved in space in a known way. By convention, let the arrangement of the projections of the selected cloud points onto the given planes be called a pattern of the intersection of the point clouds with the planes. By varying the displacement of the set of planes, maximize the coincidence of the intersection patterns of the first and second clouds of points. The degree of coincidence of the intersection patterns is characterized by a specified objective function whose arguments are target pose parameters. The target displacement parameters are found as the arguments of the objective function that maximize it. A variant of the objective function is proposed. A test example of the application of the proposed method is considered. A global optimization method known as the direct search method was used to find the arguments of the objective function. Test calculations were performed, and they confirmed the workability of the algorithm and determined its scope of applicability. The computational burden was not considered: the goal was to verify the workability of the method in principle. The advantage of the proposed method is its ability to determine a change in the pose of an object with an unknown surface shape. The proposed method may be considered as a source of observation data for a filtering procedure when estimating the parameters of the relative motion of a target with an unknown surface shape.

DeepArUco++: Improved detection of square fiducial markers in challenging lighting conditions

Fiducial markers are a computer vision tool used for object pose estimation and detection. These markers are highly useful in fields such as industry, medicine and logistics. However, optimal lighting conditions are not always available, and other factors such as blur or sensor noise can affect image quality. Classical computer vision techniques that precisely locate and decode fiducial markers often fail under difficult illumination conditions (e.g. extreme variations of lighting within the same frame). Hence, we propose DeepArUco++, a deep learning-based framework that leverages the robustness of Convolutional Neural Networks to perform marker detection and decoding in challenging lighting conditions. The framework is based on a pipeline using different Neural Network models at each step, namely marker detection, corner refinement and marker decoding. Additionally, we propose a simple method for generating synthetic data for training the different models that compose the proposed pipeline, and we present a second, real-life dataset of ArUco markers in challenging lighting conditions used to evaluate our system. The developed method outperforms other state-of-the-art methods in such tasks and remains competitive even when testing on the datasets used to develop those methods. Code available in GitHub: https://github.com/AVAuco/deeparuco/.

Dexterous Manipulation Based on Object Recognition and Accurate Pose Estimation Using RGB-D Data.

This study presents an integrated system for object recognition, six-degrees-of-freedom pose estimation, and dexterous manipulation using a JACO robotic arm with an Intel RealSense D435 camera. This system is designed to automate the manipulation of industrial valves by capturing point clouds (PCs) from multiple perspectives to improve the accuracy of pose estimation. The object recognition module includes scene segmentation, geometric primitives recognition, model recognition, and a color-based clustering and integration approach enhanced by a dynamic cluster merging algorithm. Pose estimation is achieved using the random sample consensus algorithm, which predicts position and orientation. The system was tested within a 60° field of view, which extended in all directions in front of the object. The experimental results show that the system performs reliably within acceptable error thresholds for both position and orientation when the objects are within a ±15° range of the camera's direct view. However, errors increased with more extreme object orientations and distances, particularly when estimating the orientation of ball valves. A zone-based dexterous manipulation strategy was developed to overcome these challenges, where the system adjusts the camera position for optimal conditions. This approach mitigates larger errors in difficult scenarios, enhancing overall system reliability. The key contributions of this research include a novel method for improving object recognition and pose estimation, a technique for increasing the accuracy of pose estimation, and the development of a robot motion model for dexterous manipulation in industrial settings.

Efficient object recognition under cluttered scenes via descriptor-based matching and single point voting

This paper addresses the problem of recognizing multiple objects and multiple instances from point clouds. Whereas existing methods utilize descriptors on 3D fields or pointwise voting to achieve this task, our framework takes advantage of both descriptor-based and voting-based schemes to realize more robust and efficient prediction. Specifically, we propose a novel and robust descriptor called an orientation-enhanced fast point feature histogram (OE-FPFH) to describe points in both the object model and scene, and further to build the correspondence set. The OE-FPFH integrates an orientation vector through mining the geometric tensor of the local structure of a surface point, which is more representative than the original FPFH descriptor. To improve voting efficiency, we devise a novel single-point voting mechanism (SPVM), which constructs a unique local reference frame (LRF) on a single point using the orientation vector. The SPVM takes as input the corresponding point set and can generate a pose candidate for each correspondence. The process is realized by matching LRFs from two corresponding points. All pose candidates are subsequently divided into clusters and aggregated using the K-means clustering algorithm to deduce the poses for different objects or instances in the scene. Experiments on three challenging datasets demonstrate that our method is effective, efficient, and robust to occlusions and multiple instances.

Multi-Modal Pose Representations for 6-DOF Object Tracking

Pose estimation methods for robotics should return a distribution of poses rather than just a single pose estimate. Motivated by this, in this work we investigate multi-modal pose representations for reliable 6-DoF object tracking. A neural network architecture for simultaneous object segmentation and estimation of fiducial points of the object on RGB images is proposed. Given a priori probability distribution of object poses a particle filter is employed to estimate the posterior probability distribution of object poses. An advanced observation model relying on matching the projected 3D model with the segmented object and a distance transform-based object representation is used to weight samples representing the probability distribution. Afterwards, the object pose determined by the PnP algorithm is included in the probability distribution via replacing a particle with the smallest weight. Next, a k-means++ algorithm is executed to determine modes in a multi-modal probability distribution. A multi-swarm particle swarm optimization is then executed to determine the finest modes in the probability distribution. A subset of particles for final pose optimization is found in a multi-criteria analysis using the TOPSIS algorithm. They are verified using conflicting criteria that are determined on the basis of object keypoints, segmented object, and the distance transform. On the challenging YCB-Video dataset it outperforms recent algorithms for both object pose estimation and object pose tracking.

SLAM[formula omitted]: Simultaneous Localization and Multimode Mapping for indoor dynamic environments

Traditional visual Simultaneous Localization and Mapping (SLAM) methods based on point features are often limited by strong static assumptions and texture information, resulting in inaccurate camera pose estimation and object localization. To address these challenges, we present SLAM2, a novel semantic RGB-D SLAM system that can obtain accurate estimation of the camera pose and the 6DOF pose of other objects, resulting in complete and clean static 3D model mapping in dynamic environments. Our system makes full use of the point, line, and plane features in space to enhance the camera pose estimation accuracy. It combines the traditional geometric method with a deep learning method to detect both known and unknown dynamic objects in the scene. Moreover, our system is designed with a three-mode mapping method, including dense, semi-dense, and sparse, where the mode can be selected according to the needs of different tasks. This makes our visual SLAM system applicable to diverse application areas. Evaluation in the TUM RGB-D and Bonn RGB-D datasets demonstrates that our SLAM system achieves the most advanced localization accuracy and the cleanest static 3D mapping of the scene in dynamic environments, compared to state-of-the-art methods. Specifically, our system achieves a root mean square error (RMSE) of 0.018 m in the highly dynamic TUM w/half sequence, outperforming ORB-SLAM3 (0.231 m) and DRG-SLAM (0.025 m). In the Bonn dataset, our system demonstrates superior performance in 14 out of 18 sequences, with an average RMSE reduction of 27.3% compared to the next best method.

Research on the Cable-to-Terminal Connection Recognition Based on the YOLOv8-Pose Estimation Model

Substations, as critical nodes for power transmission and distribution, play a pivotal role in ensuring the stability and security of the entire power grid. With the ever-increasing demand for electricity and the growing complexity of grid structures, traditional manual inspection methods for substations can no longer meet the requirements for efficient and safe operation and maintenance. The advent of automated inspection systems has brought revolutionary changes to the power industry. These systems utilize advanced sensor technology, image processing techniques, and artificial intelligence algorithms to achieve real-time monitoring and fault diagnosis of substation equipment. Among these, the recognition of cable-to-terminal connection relationships is a key task for automated inspection systems, and its accuracy directly impacts the system’s diagnostic capabilities and fault prevention levels. However, traditional methods face numerous limitations when dealing with complex power environments, such as inadequate recognition performance under conditions of significant perspective angles and geometric distortions. This paper proposes a cable-to-terminal connection relationship recognition method based on the YOLOv8-pose model. The YOLOv8-pose model combines object detection and pose estimation techniques, significantly improving detection accuracy and real-time performance in environments with small targets and dense occlusions through optimized feature extraction algorithms and enhanced receptive fields. The model achieves an average inference time of 74 milliseconds on the test set, with an accuracy of 92.8%, a recall rate of 91.5%, and an average precision mean of 90.2%. Experimental results demonstrate that the YOLOv8-pose model performs excellently under different angles and complex backgrounds, accurately identifying the connection relationships between terminals and cables, providing reliable technical support for automated substation inspection systems. This research offers an innovative solution for automated substation inspection systems, with significant application prospects.

FusionNetV2: Explicit Enhancement of Edge Features for 6D Object Pose Estimation

FusionNet is a hybrid model that incorporates convolutional neural networks and Transformers, achieving state-of-the-art performance in 6D object pose estimation while significantly reducing the number of model parameters. Our study reveals that FusionNet has local and global attention mechanisms for enhancing deep features in two paths and the attention mechanisms play a role in implicitly enhancing features around object edges. We found that enhancing the features around object edges was the main reason for the performance improvement in 6D object pose estimation. Therefore, in this study, we attempt to enhance the features around object edges explicitly and intuitively. To this end, an edge boosting block (EBB) is introduced that replaces the attention blocks responsible for local attention in FusionNet. EBB is lightweight and can be directly applied to FusionNet with minimal modifications. EBB significantly improved the performance of FusionNet in 6D object pose estimation in experiments on the LINEMOD dataset.

Six-Degree-of-Freedom Pose Estimation Method for Multi-Source Feature Points Based on Fully Convolutional Neural Network

An object’s six-degree-of-freedom (6DoF) pose information has great importance in various fields. Existing methods of pose estimation usually detect two-dimensional (2D)-three-dimensional (3D) feature point pairs, and directly estimates the pose information through Perspective-n-Point (PnP) algorithms. However, this approach ignores the spatial association between pixels, making it difficult to obtain high-precision results. In order to apply pose estimation based on deep learning methods to real-world scenarios, we hope to design a method that is robust enough in more complex scenarios. Therefore, we introduce a method for 3D object pose estimation from color images based on farthest point sampling (FPS) and object 3D bounding box. This method detects the 2D projection of 3D feature points through a convolutional neural network, matches it with the 3D model of the object, and then uses the PnP algorithm to restore the feature point pair to the object pose. Due to the global nature of the bounding box, this approach can be considered effective even in partially occluded or complex environments. In addition, we propose a heatmap suppression method based on weighted coordinates to further improve the prediction accuracy of feature points and the accuracy of the PnP algorithm in solving the pose position. Compared with other algorithms, this method has higher accuracy and better robustness. Our method yielded 93.8% of the ADD(-s) metrics on the Linemod dataset and 47.7% of the ADD(-s) metrics on the Occlusion Linemod dataset. These results show that our method is more effective than existing methods in pose estimation of large objects.

Refined Prior Guided Category-Level 6D Pose Estimation and Its Application on Robotic Grasping

Estimating the 6D pose and size of objects is crucial in the task of visual grasping for robotic arms. Most current algorithms still require the 3D CAD model of the target object to match with the detected points, and they are unable to predict the object’s size, which significantly limits the generalizability of these methods. In this paper, we introduce category priors and extract high-dimensional abstract features from both the observed point cloud and the prior to predict the deformation matrix of the reconstructed point cloud and the dense correspondence between the reconstructed and observed point clouds. Furthermore, we propose a staged geometric correction and dense correspondence refinement mechanism to enhance the accuracy of regression. In addition, a novel lightweight attention module is introduced to further integrate the extracted features and identify potential correlations between the observed point cloud and the category prior. Ultimately, the object’s translation, rotation, and size are obtained by mapping the reconstructed point cloud to a normalized canonical coordinate system. Through extensive experiments, we demonstrate that our algorithm outperforms existing methods in terms of performance and accuracy on commonly used benchmarks for this type of problem. Additionally, we implement the algorithm in robotic arm-grasping simulations, further validating its effectiveness.

End-to-End Implicit Object Pose Estimation.

To accurately estimate the 6D pose of objects, most methods employ a two-stage algorithm. While such two-stage algorithms achieve high accuracy, they are often slow. Additionally, many approaches utilize encoding-decoding to obtain the 6D pose, with many employing bilinear sampling for decoding. However, bilinear sampling tends to sacrifice the accuracy of precise features. In our research, we propose a novel solution that utilizes implicit representation as a bridge between discrete feature maps and continuous feature maps. We represent the feature map as a coordinate field, where each coordinate pair corresponds to a feature value. These feature values are then used to estimate feature maps of arbitrary scales, replacing upsampling for decoding. We apply the proposed implicit module to a bidirectional fusion feature pyramid network. Based on this implicit module, we propose three network branches: a class estimation branch, a bounding box estimation branch, and the final pose estimation branch. For this pose estimation branch, we propose a miniature dual-stream network, which estimates object surface features and complements the relationship between 2D and 3D. We represent the rotation component using the SVD (Singular Value Decomposition) representation method, resulting in a more accurate object pose. We achieved satisfactory experimental results on the widely used 6D pose estimation benchmark dataset Linemod. This innovative approach provides a more convenient solution for 6D object pose estimation.

Deocclusion and integration of advantages for a better hand pose

Estimating hand pose in the case of hand and object interaction faces the challenge of occlusion. Traditional methods that use contact information to alleviate this problem have limited applications because accurately estimating the required shape and pose of unfamiliar objects is a difficult task. This paper address this problem with occlusion removal at image level, removing the object by the proposed self-supervision method, which reduces the labor required to collect the paired labels of occlusion and deocclusion of the hand. In addition, this paper consider occlusion as valuable information and propose an integration strategy to enrich the extracted features from the occluded hand image and deoccluded hand image. Validation experiments are performed to show the proposed model’s main contributions. The experimental results on two widely-used public datasets demonstrate that the proposed model outperforms other state-of-the-art methods.

6DoF Object Pose and Focal Length Estimation from Single RGB Images in Uncontrolled Environments.

Accurate 6DoF (degrees of freedom) pose and focal length estimation are important in extended reality (XR) applications, enabling precise object alignment and projection scaling, thereby enhancing user experiences. This study focuses on improving 6DoF pose estimation using single RGB images of unknown camera metadata. Estimating the 6DoF pose and focal length from an uncontrolled RGB image, obtained from the internet, is challenging because it often lacks crucial metadata. Existing methods such as FocalPose and Focalpose++ have made progress in this domain but still face challenges due to the projection scale ambiguity between the translation of an object along the z-axis (tz) and the camera's focal length. To overcome this, we propose a two-stage strategy that decouples the projection scaling ambiguity in the estimation of z-axis translation and focal length. In the first stage, tz is set arbitrarily, and we predict all the other pose parameters and focal length relative to the fixed tz. In the second stage, we predict the true value of tz while scaling the focal length based on the tz update. The proposed two-stage method reduces projection scale ambiguity in RGB images and improves pose estimation accuracy. The iterative update rules constrained to the first stage and tailored loss functions including Huber loss in the second stage enhance the accuracy in both 6DoF pose and focal length estimation. Experimental results using benchmark datasets show significant improvements in terms of median rotation and translation errors, as well as better projection accuracy compared to the existing state-of-the-art methods. In an evaluation across the Pix3D datasets (chair, sofa, table, and bed), the proposed two-stage method improves projection accuracy by approximately 7.19%. Additionally, the incorporation of Huber loss resulted in a significant reduction in translation and focal length errors by 20.27% and 6.65%, respectively, in comparison to the Focalpose++ method.

Category-Level Object Pose Estimation with Statistic Attention.

Six-dimensional object pose estimation is a fundamental problem in the field of computer vision. Recently, category-level object pose estimation methods based on 3D-GC have made significant breakthroughs due to advancements in 3D-GC. However, current methods often fail to capture long-range dependencies, which are crucial for modeling complex and occluded object shapes. Additionally, discerning detailed differences between different objects is essential. Some existing methods utilize self-attention mechanisms or Transformer encoder-decoder structures to address the lack of long-range dependencies, but they only focus on first-order information of features, failing to explore more complex information and neglecting detailed differences between objects. In this paper, we propose SAPENet, which follows the 3D-GC architecture but replaces the 3D-GC in the encoder part with HS-layer to extract features and incorporates statistical attention to compute higher-order statistical information. Additionally, three sub-modules are designed for pose regression, point cloud reconstruction, and bounding box voting. The pose regression module also integrates statistical attention to leverage higher-order statistical information for modeling geometric relationships and aiding regression. Experiments demonstrate that our method achieves outstanding performance, attaining an mAP of 49.5 on the 5°2 cm metric, which is 3.4 higher than the baseline model. Our method achieves state-of-the-art (SOTA) performance on the REAL275 dataset.

Recognition and pose estimation method for random bin picking considering incomplete point cloud scene

PurposeThis paper aims to propose a stable and precise recognition and pose estimation method to deal with the difficulties that industrial parts often present, such as incomplete point cloud data due to surface reflections, lack of color texture features and limited availability of effective three-dimensional geometric information. These challenges lead to less-than-ideal performance of existing object recognition and pose estimation methods based on two-dimensional images or three-dimensional point cloud features.Design/methodology/approachIn this paper, an image-guided depth map completion method is proposed to improve the algorithm's adaptability to noise and incomplete point cloud scenes. Furthermore, this paper also proposes a pose estimation method based on contour feature matching.FindingsThrough experimental testing on real-world and virtual scene dataset, it has been verified that the image-guided depth map completion method exhibits higher accuracy in estimating depth values for depth map hole pixels. The pose estimation method proposed in this paper was applied to conduct pose estimation experiments on various parts. The average recognition accuracy in real-world scenes was 88.17%, whereas in virtual scenes, the average recognition accuracy reached 95%.Originality/valueThe proposed recognition and pose estimation method can stably and precisely deal with the difficulties that industrial parts present and improve the algorithm's adaptability to noise and incomplete point cloud scenes.

Adversarial imitation learning-based network for category-level 6D object pose estimation

Adversarial imitation learning-based network for category-level 6D object pose estimation