aswinkumar99/LeRobot-SO101-task2-multiple-sponges-no-distractors-random-locations

真实世界的SO101拾取放置数据集，针对多个海绵且无干扰物的情况。此分割使用随机化的工作空间布局。数据集包含64个片段，69013帧，帧率为30fps，使用了两台摄像机（头顶和手腕），光照条件包括早晨和夜晚各32次。旨在用于模仿学习、行为克隆、策略评估和布局变化下的鲁棒性研究。局限性包括仅包含一种任务变体、仅使用海绵物体、摄像机位置固定等。

Real-world SO101 pick-and-place dataset for multiple sponges with no distractors. This split uses randomized workspace layouts. The dataset contains 64 episodes, 69013 frames, with a frame rate of 30fps, using two cameras (overhead and wrist), and lighting conditions include 32 morning and 32 night. Intended for imitation learning, behavior cloning, policy evaluation, and robustness studies under layout changes. Limitations include only one task variant, only sponge objects are used, and camera placement is fixed.