We present FuncGrasp, a framework that can infer dense yet reliable grasp
configurations for unseen objects using one annotated object and single-view
RGB-D observation via categorical priors. Unlike previous works that only
transfer a set of grasp poses, FuncGrasp aims to transfer infinite
configurations parameterized by an object-centric continuous grasp function
across varying instances. To ease the transfer process, we propose Neural
Surface Grasping Fields (NSGF), an effective neural representation defined on
the surface to densely encode grasp configurations. Further, we exploit
function-to-function transfer using sphere primitives to establish semantically
meaningful categorical correspondences, which are learned in an unsupervised
fashion without any expert knowledge. We showcase the effectiveness through
extensive experiments in both simulators and the real world. Remarkably, our
framework significantly outperforms several strong baseline methods in terms of
density and reliability for generated grasps.

FuncGrasp 是一个框架，可以通过一个带有注释的对象和单视图 RGB-D 观察来推断未见过的物体的密集而可靠的抓取配置，通过球形基元的函数到函数传递建立语义有意义的分类对应关系，并在无监督学习的方式下学习。与以往的作品不同，FuncGrasp 旨在传输通过对象为中心的连续抓握函数参数化的无限配置，通过在表面上定义的有效神经表示（NSGF）密集编码抓握配置。通过在模拟器和真实世界中进行大量实验，我们展示了该方法的有效性，显着优于生成的抓握的密度和可靠性的几种强基准方法。

FuncGrasp: 从单个标注示例对象学习基于对象的神经抓取函数

FuncGrasp: Learning Object-Centric Neural Grasp Functions from Single  Annotated Example Object

As robots become more accessible outside of industrial settings, the need for
reliable object grasping and manipulation grows significantly. In such dynamic
environments it is expected that the robot is capable of reliably grasping and
manipulating novel objects in different situations. In this work we present
GraspCaps: a novel architecture based on Capsule Networks for generating
per-point grasp configurations for familiar objects. In our work, the
activation vector of each capsule in the deepest capsule layer corresponds to
one specific class of object. This way, the network is able to extract a rich
feature vector of the objects present in the point cloud input, which is then
used for generating per-point grasp vectors. This approach should allow the
network to learn specific grasping strategies for each of the different object
categories. Along with GraspCaps we present a method for generating a large
object grasping dataset using simulated annealing. The obtained dataset is then
used to train the GraspCaps network. We performed an extensive set of
experiments to assess the performance of the proposed approach regarding
familiar object recognition accuracy and grasp success rate on challenging real
and simulated scenarios.

本文介绍了一种基于胶囊网络的 GraspCaps 架构，用于根据点云数据生成每个点的物体抓取向量，并使用模拟退火方法生成大型抓取数据集进行网络训练。作者在真实和模拟场景下进行了大量实验，评估了所提出方法的表现，包括在熟悉的对象识别准确性和抓握成功率方面的表现。