cyberorigin/cyber_take_the_item_v0.1 · Datasets at Hugging Face

token_dtype stringclasses 1 value	s int64 16 16	h int64 16 16	w int64 16 16	vocab_size int64 262k 262k	hz int64 30 30	tokenizer_ckpt stringclasses 1 value	num_images int64 218k 399k	num_episodes int64 1.07k 1.93k	task stringclasses 1 value
uint32	16	16	16	262,144	30	data/magvit2.ckpt	282,241	1,639	take_the_item
uint32	16	16	16	262,144	30	data/magvit2.ckpt	312,472	1,650	take_the_item
uint32	16	16	16	262,144	30	data/magvit2.ckpt	285,401	1,397	take_the_item
uint32	16	16	16	262,144	30	data/magvit2.ckpt	233,656	1,115	take_the_item
uint32	16	16	16	262,144	30	data/magvit2.ckpt	217,965	1,073	take_the_item
uint32	16	16	16	262,144	30	data/magvit2.ckpt	398,778	1,694	take_the_item
uint32	16	16	16	262,144	30	data/magvit2.ckpt	380,849	1,879	take_the_item
uint32	16	16	16	262,144	30	data/magvit2.ckpt	368,323	1,900	take_the_item
uint32	16	16	16	262,144	30	data/magvit2.ckpt	352,934	1,822	take_the_item
uint32	16	16	16	262,144	30	data/magvit2.ckpt	246,178	1,925	take_the_item
uint32	16	16	16	262,144	30	data/magvit2.ckpt	250,753	1,189	take_the_item
uint32	16	16	16	262,144	30	data/magvit2.ckpt	288,209	1,496	take_the_item

token_dtype

stringclasses

1 value

int64

vocab_size

int64

262k

int64

tokenizer_ckpt

stringclasses

1 value

num_images

int64

218k

399k

num_episodes

int64

1.07k

1.93k

task

stringclasses

1 value

uint32

262,144

data/magvit2.ckpt

282,241

1,639

take_the_item

uint32

262,144

data/magvit2.ckpt

312,472

1,650

take_the_item

uint32

262,144

data/magvit2.ckpt

285,401

1,397

take_the_item

uint32

262,144

data/magvit2.ckpt

233,656

1,115

take_the_item

uint32

262,144

data/magvit2.ckpt

217,965

1,073

take_the_item

uint32

262,144

data/magvit2.ckpt

398,778

1,694

take_the_item

uint32

262,144

data/magvit2.ckpt

380,849

1,879

take_the_item

uint32

262,144

data/magvit2.ckpt

368,323

1,900

take_the_item

uint32

262,144

data/magvit2.ckpt

352,934

1,822

take_the_item

uint32

262,144

data/magvit2.ckpt

246,178

1,925

take_the_item

uint32

262,144

data/magvit2.ckpt

250,753

1,189

take_the_item

uint32

262,144

data/magvit2.ckpt

288,209

1,496

take_the_item

CyberOrigin Dataset

Our data includes information from home services, the logistics industry, and laboratory scenarios. For more details, please refer to our Offical Data Website

contents of dataset:

cyber_take_the_item # dataset root path
  └── data/
      ├── metadata_ID1_240808.json
      ├── segment_ids_ID1_240808.bin # for each frame segment_ids uniquely points to the segment index that frame i came from. You may want to use this to separate non-contiguous frames from different videos (transitions).
      ├── videos_ID1_240808.bin # 16x16 image patches at 30hz, each patch is vector-quantized into 2^18 possible integer values. These can be decoded into 256x256 RGB images using the provided magvit2.ckpt weights.
      ├── ...
  └── ...

{
    "task": "Take the Item",
    "total_episodes": 18779,
    "total_frames": 11982851,
    "token_dtype": "uint32",
    "vocab_size": 262144,
    "fps": 30,
    "manipulation_type": "Single Arm",
    "language_annotation": "None",
    "scene_type": "Table Top",
    "data_collect_method": "Directly Collection on Human"
}