etri-robotai-platform-dev/pusht-diffusion-policy-baseline-v1
收藏数据链接：
官方服务：
资源简介：
Push-T是Diffusion Policy论文中的代表性基准任务，涉及平面推杆将T形块对齐到目标位置。数据集来自lerobot/pusht，总共有206个episode，选取前100个用于训练，以缩短回归和smoke测试周期。推荐用于Diffusion Policy/VQ-BeT回归学习，评估通过pusht-sim进行（成功率和平均IoU）。
Push-T is a representative baseline task in the Diffusion Policy paper, where a planar pusher aligns a T-shaped block to a target position. The dataset is sourced from lerobot/pusht, with a total of 206 episodes, of which the first 100 are selected for training to shorten the regression and smoke test cycle. It is recommended for Diffusion Policy/VQ-BeT regression learning, evaluated via pusht-sim (success rate and average IoU).
