Dataset Viewer
Auto-converted to Parquet Duplicate
name
string
description
string
version
string
is_public
bool
created_at
string
owner_id
string
filters
dict
sources
list
total_instances
int64
selected_indices_source
string
Push-T Diffusion Policy baseline
Diffusion Policy 논문 baseline 으로 표준화된 Push-T task. 학습용 첫 100 에피소드 (전체 206 중) 만 사용해 빠른 회귀 검증.
1.0
true
2026-05-26T07:50:48.950543+00:00
bf95d70f-e08e-4dfd-b592-c68d40851834
{ "task": [ "push_t" ], "benchmark": [ "diffusion_policy_2024" ], "robot_family": [ "planar_pusher" ], "episode_range": [ "head_100" ] }
[ { "dataset_id": "037eacf2-a321-4457-b9ae-354c4f4c482e", "hf_repo_id": "lerobot/pusht", "commit_hash": "7628202a2180972f291ba1bc6723834921e72c19", "selected_indices": [ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, ...
100
client_provided

YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

Push-T Diffusion Policy baseline

Push-T 는 Diffusion Policy 논문의 대표 baseline task 로, planar pusher 가 T 모양 블록을 목표 위치로 정렬합니다. 전체 206 에피소드 중 학습용 첫 100 만 선별하여 회귀 + smoke 테스트 사이클을 단축했습니다.

  • 권장 용도: Diffusion Policy / VQ-BeT 회귀 학습
  • 평가 트랙: pusht-sim (성공률 + 평균 IoU)

Sources

Dataset Source Total Selected
lerobot-pusht lerobot/pusht 206 100

Filters

{
  "task": [
    "push_t"
  ],
  "benchmark": [
    "diffusion_policy_2024"
  ],
  "robot_family": [
    "planar_pusher"
  ],
  "episode_range": [
    "head_100"
  ]
}

License & Provenance

  • 본 큐레이션은 dev 검증용으로 ETRI RobotAI 플랫폼에서 자동 생성되었습니다.
  • 각 source 데이터셋의 원본 라이선스를 따릅니다.
  • manifest.jsonselected_indices 가 실제 학습/평가에 사용할 episode index 입니다.
Downloads last month
33