The RU-APC (Rutgers APC) dataset is a valuable resource for researchers and developers working on robotic perception solutions for warehouse picking challenges. Let me provide you with some details about this dataset:
- Dataset Overview:
- The Rutgers APC RGB-D Dataset is provided by the PRACSYS lab at Rutgers University.
- It is designed to equip the research community with rich data for evaluating and improving robotic perception in the context of warehouse automation.
-
The dataset contains 10,368 depth and RGB registered images.
-
Object Annotations:
- For 24 of the Amazon Picking Challenge (APC) objects, the dataset includes hand-annotated 6DOF poses.
-
These annotations are crucial for accurate pose estimation during object manipulation tasks.
-
3D Mesh Models:
- Alongside the images, the dataset provides 3D mesh models for all 25 APC objects (excluding the mead_index_cards).
-
These mesh models can be used for training recognition algorithms.
-
Context and Significance:
- The Amazon Picking Challenge (APC) is a competition that focuses on perception, motion planning, and grasping of different objects placed inside bins of an Amazon-Kiva Pod.
- Warehouse automation, especially picking and placing products on shelves, is a critical area of interest.
- Unlike simpler tabletop scenarios, warehouse shelves introduce challenges due to narrow, dark, and obscuring bins.
-
Accurate pose estimation is essential for successful object manipulation within these shelves.
-
Variety of Objects:
- The selected objects in the dataset were used during the first Amazon Picking Challenge held in Seattle in May 2015.
- These objects exhibit diversity in terms of size, shape, texture, transparency, and other characteristics.
- They represent good candidates for robotic units to transport in warehouse environments.
(1) Rutgers APC RGB-D Dataset - PRACSYS Group. https://robotics.cs.rutgers.edu/pracsys/rutgers-apc-rgb-d-dataset/.
(2) Datasets - PRACSYS Group. https://robotics.cs.rutgers.edu/pracsys/datasets/.
(3) Datasets - BOP: Benchmark for 6D Object Pose Estimation. https://bop.felk.cvut.cz/datasets/.