Sign in to view source links and access this dataset
Description
TaskGrasp-Pro extends the original TaskGrasp dataset with fine-grained part decompositions and part-level physical property annotations for 190 household objects. The dataset, created by WCL-Robotics, defines three task types per object instance, resulting in a total of 2,850 tasks. It includes point clouds, multi-view RGB-D images, and task-related data.
Use Cases
Train models for robotic grasping based on part-level physical property annotations.
Develop algorithms for task-oriented object segmentation using fine-grained part decompositions.
Benchmark multi-task learning systems using the defined category-related, part-related, and part-irrelevant tasks.
Generate synthetic training data for object recognition using multi-view RGB-D images and point clouds.
Strengths
Provides annotations for 190 household objects.
Defines 2,850 tasks across three distinct types.
Includes multiple data modalities such as point clouds and RGB-D images.
Limitations
Column-level documentation is absent; field semantics must be inferred after download.
Row count is unknown, which may limit suitability assessment.
Freshness should be verified as the last update date is in the future (2026-04-30).
Provenance
Source
WCL-Robotics
Freshness
Last updated 2026-04-30 08:02:15
License is unknown; terms of use must be verified before application.