Abstract:We propose a fast and accurate method of 6D object pose estimation for bin-picking of mechanical parts by a robot manipulator. We extend the single-shot approach to stereo vision by application of attention architecture. Our convolutional neural network model regresses to object locations and rotations from either a left image or a right image without depth information. Then, a stereo feature matching module, designated as Stereo Grid Attention, generates stereo grid matching maps. The important point of our method is only to calculate disparity of the objects found by the attention from stereo images, instead of calculating a point cloud over the entire image. The disparity value is then used to calculate the depth to the objects by the principle of triangulation. Our method also achieves a rapid processing speed of pose estimation by the single-shot architecture and it is possible to process a 1024 x 1024 pixels image in 75 milliseconds on the Jetson AGX Xavier implemented with half-float model. Weakly textured mechanical parts are used to exemplify the method. First, we create original synthetic datasets for training and evaluating of the proposed model. This dataset is created by capturing and rendering numerous 3D models of several types of mechanical parts in virtual space. Finally, we use a robotic manipulator with an electromagnetic gripper to pick up the mechanical parts in a cluttered state to verify the validity of our method in an actual scene. When a raw stereo image is used by the proposed method from our stereo camera to detect black steel screws, stainless screws, and DC motor parts, i.e., cases, rotor cores and commutator caps, the bin-picking tasks are successful with 76.3%, 64.0%, 50.5%, 89.1% and 64.2% probability, respectively.

Towards One Shot & Pick All: 3D-OAS, an end-to-end framework for vision guided top-down parcel bin-picking using 3D-overlapping-aware instance segmentation and GNN

Fast and Robust Bin-picking System for Densely Piled Industrial Objects

A Robotic Semantic Grasping Method for Pick-and-place Tasks

DoraPicker: An Autonomous Picking System for General Objects

Picking from Clutter: An Object Segmentation Method for Robot Grasping.

Bin-picking of novel objects through category-agnostic-segmentation: RGB matters

Visual Sorting of Express Parcels Based on Multi-Task Deep Learning

Two-Stage Grasping: A New Bin Picking Framework for Small Objects

Efficient End-to-End Detection of 6-DoF Grasps for Robotic Bin Picking

Towards Cross-device and Training-free Robotic Grasping in 3D Open World

Iterative Visual Recognition for Learning Based Randomized Bin-Picking

A Low-Cost, High-Speed, and Robust Bin Picking System for Factory Automation Enabled by a Non-Stop, Multi-View, and Active Vision Scheme

Vision-centric 3D point cloud technique and custom gripper process for parcel depalletisation

Unseen Object Few-Shot Semantic Segmentation for Robotic Grasping

Design of an Unordered Gripping System Based on 3D Vision

Robot Online 3D Bin Packing Strategy Based on Deep Reinforcement Learning and 3D Vision

Stereo Vision Based Single-Shot 6D Object Pose Estimation for Bin-Picking by a Robot Manipulator

A graph reasoning method for multi-object unordered stacking scenarios

Low Cost Bin Picking Solution for E-Commerce Warehouse Fulfillment Centers

S4G: Amodal Single-view Single-Shot SE(3) Grasp Detection in Cluttered Scenes

Revolutionizing Robotic Depalletizing: AI-Enhanced Parcel Detecting with Adaptive 3D Machine Vision and RGB-D Imaging for Automated Unloading