Single Shot 6D Object Pose Estimation

Kleeberger, Kilian; Huber, Marco F.

doi:10.1109/icra40945.2020.9197207

Cited by 33 publications

(18 citation statements)

References 33 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Considering these inputs, a broad set of approaches has been introduced. For example, the OP-Net was presented in [21]. The OP-net is a fully convolutional neural network trained on synthetic data.…”

Section: A Deep Convolutional Network For Object Position and Pose Estimationmentioning

confidence: 99%

Memory Efficient Grasping Point Detection of Nontrivial Objects

et al. 2021

View full text Add to dashboard Cite

Robotic manipulation with a nontrivial object providing various types of grasping points is of an industrial interest. Here, an efficient method of simultaneous detection of the grasping points is proposed. Specifically, two different 3 degree-of-freedom end effectors are considered for simultaneous grasping. The method utilizes an RGB data-driven perception system based on a specifically designed fully convolutional neural network called attention squeeze parallel U-Net (ASP U-Net). ASP U-Net detects grasping points based on a single RGB image. This image is transformed into a schematic grayscale frame, where the positions and poses of the grasping points are coded into gradient geometric shapes. In order to approve the ASP U-Net architecture, its performance was compared with nine competitive architectures using metrics based on generalized intersection over union and mean absolute error. The results indicate its outstanding accuracy and response time. ASP U-Net is also computationally efficient enough. With a more than acceptable memory size (77 MB), the architecture can be implemented using custom single-board computers. Here, its capabilities were tested and evaluated on the NVIDIA Jetson NANO platform.INDEX TERMS Robotic grasping, grasping point detection, machine vision, deep learning, convolutional neural network.

show abstract

Section: A Deep Convolutional Network For Object Position and Pose Estimationmentioning

confidence: 99%

Memory Efficient Grasping Point Detection of Nontrivial Objects

et al. 2021

View full text Add to dashboard Cite

show abstract

“…Early approaches were based on aligning 3D CAD models to 3D point clouds using hand-crafted features [4,5] and variants of iterative closest point algorithm [6], using either depth or RGB-D sensors [7]. Recently, CNN-based deep learning algorithms have been developed to estimate the 6 dof pose of an object in a single shot [2,8,9,10]. For example, PoseNet [11] proposed a CNN model to regress a 6 dof camera pose from a single RGB image, whereas PoseCNN [12] localized objects in the 2D image and predicted their depth information to generate 3D locations.…”

Section: Related Workmentioning

confidence: 99%

“…Since our proposed method accepts object poses as input, we need a pose estimation network to train and evaluate the effectiveness of our method. For the estimation of 6 dof object pose from the scene, we have used OP-Net [2] proposed by Kleeberger and Huber. OP-Net was developed to address the problem of pose determination for the robotic bin-picking application, where a scene comprises multiple instances of a single known object class, as shown in Fig.…”

Section: Object Pose Estimationmentioning

confidence: 99%

“…For training OP-Net author proposed two different loss function. Here, we used L ori1 loss as it was providing better results for most of the objects in the Siléane dataset [2].…”

Section: Object Pose Estimationmentioning

confidence: 99%

“…Mazumder et al [1] proposed ValidNet which used a two-class classifier to discriminate true positive and false positive instances in 3D surface registration algorithms. OP-Net [2] proposed a 3 dof Thanks to Bluewrist Inc. and NSERC for their support of this work. confidence score (only for the translational pose component) for differentiating true and false object detections.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Multistream Validnet: Improving 6D Object Pose Estimation by Automatic Multistream Validation

Mazumder

Zand

Greenspan

2021

2021 IEEE International Conference on Image Processing (ICIP)

View full text Add to dashboard Cite

This work presents a novel approach to improve the results of pose estimation by detecting and distinguishing between the occurrence of True and False Positive results. It achieves this by training a binary classifier on the output of an arbitrary pose estimation algorithm, and returns a binary label indicating the validity of the result. We demonstrate that our approach improves upon a state-of-the-art pose estimation result on the Siléane dataset, outperforming a variation of the alternative CullNet method by 4.15% in average class accuracy and 0.73% in overall accuracy at validation. Applying our method can also improve the pose estimation average precision results of Op-Net by 6.06% on average.

show abstract