bytedance_robot_benchmark_20
收藏资源简介:
ByteDance Robot Benchmark (BDRBench-20) 是一个视觉-语言-动作(VLA)数据集,包含8K高质量轨迹,旨在评估VLA模型在现实场景中的性能。该数据集包括20个常见操作任务,如拾取和放置、倒水和开/关动作。数据集分为训练集和验证集,包含注释文件和视频文件。注释文件描述了任务、视频路径、动作和状态信息,视频文件记录了任务的执行过程。数据集设计用于训练和评估VLA模型在现实场景中的应用。
ByteDance Robot Benchmark (BDRBench-20) is a vision-language-action (VLA) dataset containing 8K high-quality trajectories, designed to evaluate the performance of VLA models in real-world scenarios. It includes 20 common manipulation tasks such as pick-and-place, pouring, and on/off operations. The dataset is divided into training and validation sets, and contains annotation files and video files. The annotation files describe task details, video paths, actions and state information, while the video files record the execution process of each task. This dataset is intended for training and evaluating the real-world application of VLA models.




