Fraunhofer IPA Bin-Picking

Fraunhofer IPA Bin-Picking 数据集是一个大规模数据集，包含各种对象（可能具有对称性）的模拟场景和真实场景，并使用 6D 姿势进行了完整注释。物理模拟用于通过将对象放置在垃圾箱上方的随机位置和方向来批量创建许多部分的场景。此外，该数据集通过提供更多样本扩展了 Siléane 数据集。这允许例如训练深度神经网络并在公共 Siléane 数据集上对性能进行基准测试。通过贡献这个数据集，我们的目标是利用深度学习等先进的机器学习技术来推进对象姿态估计新方法的开发。

The Fraunhofer IPA Bin-Picking Dataset is a large-scale dataset that encompasses both simulated and real-world scenes with various objects (which may exhibit symmetry), fully annotated with 6D poses. Physical simulations are employed to bulk-generate numerous partial scenes by placing objects at random positions and orientations above bins. Furthermore, this dataset expands the Siléane Dataset by providing additional samples. This enables, for instance, training deep neural networks and benchmarking their performance on the public Siléane Dataset. By contributing this dataset, our objective is to advance the development of novel object pose estimation methods by leveraging advanced machine learning techniques such as deep learning.

搜集汇总

背景与挑战

背景概述

Fraunhofer IPA Bin-Picking数据集是一个大规模工业场景数据集，专注于6D物体姿态估计，包含模拟和真实场景数据，并扩展了Siléane数据集以支持深度学习方法的研究和开发。