EgoDex-PickPlace-10hr

A 10-hour, lossless, per-episode subset of pick-and-place manipulation from griffinlabs/EgoDex-LeRobot-v3.0, sampled equally across 9 object categories (~1.11 h each).

Summary

• Total: 8,844 episodes / 1,080,723 frames / 10.01 hours @ 30 fps
• Format: LeRobot v3.0 (one folder per object category, each a standalone loadable dataset)
• Modality: egocentric RGB video (1920x1080, AV1) + full-body/hand pose state (no action head; EgoDex is state-only)
• Task family: basic_pick_place — Pick up the from the and place it in/on the (+ puzzle-assembly variant)

  Extraction guarantees

  • Lossless video: source AV1 is all-intra (every frame a keyframe), so per-episode clips are cut with ffmpeg stream-copy (-c copy) — bit-exact, no re-encode (verified via raw-frame MD5 match).
  • Per-episode integrity: every episode is its own .parquet + .mp4; validated meta_length == video_frames == data_rows for all episodes.
  • Equal sampling: each category seeded-shuffled and truncated to ~1.11 h (10 h / 9). Selected episodes are 100% complete (no frame dropping within an episode).

  Categories

  Category	Episodes	Hours	# distinct objects
  puzzle	158	1.115	1
  plush_toys	1,123	1.114	43
  dice_balls	1,032	1.114	16
  food	1,007	1.113	49
  stationery	918	1.112	21
  tools	1,088	1.112	26
  blocks_shapes	1,259	1.111	72
  household	1,097	1.111	26
  electronics	1,162	1.106	15
  TOTAL	8,844	10.01	—

  Objects per category (top 15 by episode count)

  puzzle — 158 eps, 1.115 h, 1 objects

  puzzle pieces (158)

  plush_toys — 1123 eps, 1.114 h, 43 objects

  yellow duck (125), gray duck (71), brown duck (55), yellow hippo (53), white duck (52), pink duck (49), yellow plushie (44), white plushie (42), blue plushie (42), brown plushie (38), pink plushie (37), gray plushie (35), purple duck (34), orange plushie (29), yellow fish (27)

  dice_balls — 1032 eps, 1.114 h, 16 objects

  green dice (206), blue dice (186), yellow dice (116), blue dice box (93), orange ball (81), tennis ball (66), blue container of slime (47), red ball (43), yellow die (38), blue butter slime (29), rubix (28), blue slime (27), white dice (26), green tennis ball (25), light blue slime box (20)

  food — 1007 eps, 1.113 h, 49 objects

  brown egg (95), red strawberry (76), yellow banana (71), pink donut (46), white egg (43), tan baguette (39), brown donut (37), brown bread (35), croissant (31), mango (29), lime (26), bread slice (25), green apple (22), yellow bread (22), red cherry (22)

  stationery — 918 eps, 1.112 h, 21 objects

  black stapler (93), black and yellow tape measure (90), purple sticky notes (68), clear safety glasses (64), black pen (58), white and red box of tape (57), red tape dispenser (52), white box of tape (46), transparent roll of tape (44), red sharpie (40), orange crayon (38), white box of staples (38), blue pen (37), brown ruler (36), yellow soap dispenser (35)

  tools — 1088 eps, 1.112 h, 26 objects

  black screwdriver (76), blue tongs (74), gray screw (69), white ring (67), black tube (56), blue brush (52), silver padlock (49), blue cooking tong (48), gray brush (46), red screwdriver (45), screwdriver (42), yellow screw (42), red and beige screw (39), black and blue saw (39), beige and red screw (38)

  blocks_shapes — 1259 eps, 1.111 h, 72 objects

  green block (98), ice cube (57), yellow block (52), wooden block (51), jenga block (50), blue lego (46), green toy block (45), wood block (45), white block (31), blue lego block (29), black wheel (28), white lego brick (27), blue block (24), white domino (23), domino (19)

  household — 1097 eps, 1.111 h, 26 objects

  yellow cup (78), sanitizer (73), clear hand sanitizer bottle (68), blue cup (58), metal lid (58), clear jar (53), clear tupperware (52), pink tea cup (50), pink plate (49), purple teacup (46), pink teacup (45), orange lip balm (44), teacup (42), thera tears (41), small cup (40)

  electronics — 1162 eps, 1.106 h, 15 objects

  grey remote (155), white airpods case (127), black mouse (126), silver iphone (109), grey iphone (95), white laptop charger (91), white usb (76), black computer mouse (75), black usb dongle (63), airpod case (56), white remote (50), pink iphone (41), iphone (41), white airpods (31), white airpod case (26)

  Load example

  from lerobot.datasets.lerobot_dataset import LeRobotDataset
  ds = LeRobotDataset("Kavin60606/EgoDex-PickPlace-10hr", root="food")  # any category folder
  print(ds.num_episodes, ds.num_frames)
  

  Derived from EgoDex-LeRobot-v3.0. Categorization via task-string object parsing + keyword grouping.