Datasets:

capstone-vlavla
/

hot3d-lapa

Modalities:

Dataset Preview

Split (1)

train

The full dataset viewer is not available (click to read why). Only showing a preview of the rows.

The dataset generation failed

Error code:   DatasetGenerationError
Exception:    ValueError
Message:      Expected object or value
Traceback:    Traceback (most recent call last):
                File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1827, in _prepare_split_single
                  for key, table in generator:
                                    ^^^^^^^^^
                File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 613, in wrapped
                  for item in generator(*args, **kwargs):
                              ~~~~~~~~~^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 281, in _generate_tables
                  examples = [ujson_loads(line) for line in batch.splitlines()]
                              ~~~~~~~~~~~^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/utils/json.py", line 20, in ujson_loads
                  return pd.io.json.ujson_loads(*args, **kwargs)
                         ~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^
              ValueError: Expected object or value
              
              The above exception was the direct cause of the following exception:
              
              Traceback (most recent call last):
                File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 1369, in compute_config_parquet_and_info_response
                  parquet_operations, partial, estimated_dataset_info = stream_convert_to_parquet(
                                                                        ~~~~~~~~~~~~~~~~~~~~~~~~~^
                      builder, max_dataset_size_bytes=max_dataset_size_bytes
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                  )
                  ^
                File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 948, in stream_convert_to_parquet
                  builder._prepare_split(split_generator=splits_generators[split], file_format="parquet")
                  ~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1694, in _prepare_split
                  for job_id, done, content in self._prepare_split_single(
                                               ~~~~~~~~~~~~~~~~~~~~~~~~~~^
                      gen_kwargs=gen_kwargs, job_id=job_id, **_prepare_split_args
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                  ):
                  ^
                File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1879, in _prepare_split_single
                  raise DatasetGenerationError("An error occurred while generating the dataset") from e
              datasets.exceptions.DatasetGenerationError: An error occurred while generating the dataset

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

End of preview.

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

Check out the documentation for more information.

HOT3D → LAPA 학습용 데이터 (Stage A)

KSC2026 세부주제2 (Hand-centric vs Object-centric) 실험용 데이터입니다. HOT3D-Clips Aria 클립에서 물체 궤적(O)과 손목 궤적(H-nogrip)을 같은 구간·프레임 기준으로 추출한 LAPA 입력 JSON입니다.

HOT3D 원본 이미지가 포함되어 있으므로 비공개 저장소로 유지하고, 데이터셋 라이선스를 따릅니다.

구성

폴더/파일	내용	클립	step
`hot3d_50/`	최초 50개 (사람이 영상으로 지시문 확정)	50	838
`hot3d_290/`	새로 샘플링한 290개 (지시문 GPT-4o 생성)	290	5,205
`hot3d_340/`	위 두 개를 합친 것	340	6,043
`hot3d_deliver.zip`	50개 이미지. 안의 JSON은 임시 지시문 버전이므로 사용하지 말 것	50	838
`images_290.zip`	290개 이미지	290	5,205
`instructions_50.csv`, `instructions_290.csv`	클립별 지시문
`summary_290.csv`	290개 클립 요약 (물체, 손, 구간 등). instruction 열은 임시 문구이므로 `instructions_290.csv`를 사용

각 폴더에 hot3d_O.json(물체 6칸), hot3d_Hnogrip.json(손목 6칸)이 있습니다.

이미지 배치

JSON의 이미지 경로는 data/hot3d/<clip>/step_XXX.jpg 입니다. hot3d_deliver.zip과 images_290.zip을 풀면 images/ 폴더가 나오고, 그 안의 내용을 LAPA/data/hot3d/에 둡니다. 클립 이름이 겹치지 않으므로 두 zip을 같은 폴더에 풀면 340개 전체가 됩니다.

사용할 데이터	이미지	JSON
50개	`hot3d_deliver.zip`	`hot3d_50/`
290개	`images_290.zip`	`hot3d_290/`
340개	두 zip 모두	`hot3d_340/`

형식

LLaVA식 JSON 리스트. 항목: id, image, conversations
raw_actions: 6칸 [dx, dy, dz, droll, dpitch, dyaw] (m, rad)
좌표계: 클립 첫 프레임 Aria RGB(214-1) 카메라 좌표계, 5Hz, 가우시안 smoothing(sigma 2.0)
step k = (프레임 start+6k의 이미지, 그 프레임에서 +6프레임까지의 변화량)
물체가 움직이는 구간만 사용 (앞 2칸 / 뒤 1칸 여유)
같은 클립에서 O와 H-nogrip은 같은 구간, 같은 프레임, 같은 지시문
지시문은 conversations[0].value의 백틱(`) 사이에 들어 있음
그리퍼 칸 없음 (H 그리퍼 라벨은 시험했으나 신뢰할 수 없어 제외)

지시문

50개: 정연 초안을 채원이 영상으로 확인해 25개 수정
290개: GPT-4o(비전, 프레임 5장)로 생성. 물체 이름은 사람이 쓰는 이름으로 매핑해 입력함. 무작위 30개 표본 검수 결과: [ 채원 확인 결과 기입 ]
클립당 지시문 1개, 위치(O/H) 무관하게 동일

사용 시 주의

5Hz 한 step 변화량이 simpler 데이터보다 큼 (위치 약 4~~6배, 회전 약 3~~12배, p99, 50개 기준)
finetune_preprocess.py는 bin을 데이터셋마다 다시 계산하므로, 50개 / 290개 / 340개 데이터는 같은 토큰이 서로 다른 값을 뜻함
--discretize_bins 245 사용 (256이면 action_vocab_size=245를 넘음)
조작 물체·구간·손은 규칙 기반 자동 선택. 양손 모호 클립이 있음 (50개 중 10개, 290개 중 24개: hand_gap < 0.1 기준)
회전 변화량은 카메라 좌표계 기준 (로봇 베이스 좌표계와 다름)
290개 처리에서 제외된 클립 10개: 조작된 물체 없음 (5mm/step 초과 이동 2 step 미만 또는 총 경로 5cm 미만)

샘플링

전체 train_aria 1,516개 클립 중 50개(clip-001849~001898)는 최초 배치로 사용
나머지 1,466개를 300개 구간으로 나눠 구간마다 무작위 1개 선택 (시드 2026), 10개 제외 후 290개 사용

담당: 데이터 변환 [ 이름 ]

Downloads last month: 23

Total file size:

2.32 GB

id string	image string	conversations list
data/hot3d/clip-001899/step_000	data/hot3d/clip-001899/step_000.jpg	[ "{\"from\":\"human\",\"value\":\"<image>\\nWhat action should the robot take to `put the glasses on the table`\"}", "{\"from\":\"gpt\",\"raw_actions\":[0.0132037935,-0.0394101201,0.0288015531,0.241830573,-0.1365528553,0.0276603099]}" ]
data/hot3d/clip-001899/step_001	data/hot3d/clip-001899/step_001.jpg	[ "{\"from\":\"human\",\"value\":\"<image>\\nWhat action should the robot take to `put the glasses on the table`\"}", "{\"from\":\"gpt\",\"raw_actions\":[0.0180850794,-0.0851255592,0.0194168391,0.27980631,-0.1236175751,0.0763251315]}" ]
data/hot3d/clip-001899/step_002	data/hot3d/clip-001899/step_002.jpg	[ "{\"from\":\"human\",\"value\":\"<image>\\nWhat action should the robot take to `put the glasses on the table`\"}", "{\"from\":\"gpt\",\"raw_actions\":[0.0162018993,-0.0803267648,-0.0057059942,0.1780120548,-0.0443044722,0.0872109123]}" ]
data/hot3d/clip-001899/step_003	data/hot3d/clip-001899/step_003.jpg	[ "{\"from\":\"human\",\"value\":\"<image>\\nWhat action should the robot take to `put the glasses on the table`\"}", "{\"from\":\"gpt\",\"raw_actions\":[0.0074261081,-0.0073248969,-0.0256740151,-0.0344865874,0.0182858985,0.1113574842]}" ]
data/hot3d/clip-001899/step_004	data/hot3d/clip-001899/step_004.jpg	[ "{\"from\":\"human\",\"value\":\"<image>\\nWhat action should the robot take to `put the glasses on the table`\"}", "{\"from\":\"gpt\",\"raw_actions\":[0.0129235208,0.0209340077,-0.0035309106,-0.2752473298,0.2019038135,0.286724108]}" ]
data/hot3d/clip-001899/step_005	data/hot3d/clip-001899/step_005.jpg	[ "{\"from\":\"human\",\"value\":\"<image>\\nWhat action should the robot take to `put the glasses on the table`\"}", "{\"from\":\"gpt\",\"raw_actions\":[-0.0064856785,0.0160841969,0.0165793967,-0.1195971356,0.161901702,0.1193368496]}" ]
data/hot3d/clip-001899/step_006	data/hot3d/clip-001899/step_006.jpg	[ "{\"from\":\"human\",\"value\":\"<image>\\nWhat action should the robot take to `put the glasses on the table`\"}", "{\"from\":\"gpt\",\"raw_actions\":[0.0070804168,0.007675824,0.0504597086,0.0623978303,0.0061063974,0.0799739802]}" ]
data/hot3d/clip-001899/step_007	data/hot3d/clip-001899/step_007.jpg	[ "{\"from\":\"human\",\"value\":\"<image>\\nWhat action should the robot take to `put the glasses on the table`\"}", "{\"from\":\"gpt\",\"raw_actions\":[0.0153145703,-0.0067628252,0.0816735977,0.0138325999,0.0619257505,0.0233931092]}" ]
data/hot3d/clip-001899/step_008	data/hot3d/clip-001899/step_008.jpg	[ "{\"from\":\"human\",\"value\":\"<image>\\nWhat action should the robot take to `put the glasses on the table`\"}", "{\"from\":\"gpt\",\"raw_actions\":[0.0133242643,-0.0141576974,0.0670862034,0.0322391941,0.0280746608,-0.0366283174]}" ]
data/hot3d/clip-001899/step_009	data/hot3d/clip-001899/step_009.jpg	[ "{\"from\":\"human\",\"value\":\"<image>\\nWhat action should the robot take to `put the glasses on the table`\"}", "{\"from\":\"gpt\",\"raw_actions\":[0.0055482604,-0.0070285734,-0.0174713106,-0.0026646612,-0.088596765,-0.0684678556]}" ]
data/hot3d/clip-001899/step_010	data/hot3d/clip-001899/step_010.jpg	[ "{\"from\":\"human\",\"value\":\"<image>\\nWhat action should the robot take to `put the glasses on the table`\"}", "{\"from\":\"gpt\",\"raw_actions\":[0.0147161799,0.0028101633,-0.1084929097,0.0025362487,-0.263476426,-0.1400298532]}" ]
data/hot3d/clip-001899/step_011	data/hot3d/clip-001899/step_011.jpg	[ "{\"from\":\"human\",\"value\":\"<image>\\nWhat action should the robot take to `put the glasses on the table`\"}", "{\"from\":\"gpt\",\"raw_actions\":[0.0275644158,0.0112490939,-0.124333943,-0.0271316001,-0.2793062982,-0.0895563526]}" ]
data/hot3d/clip-001899/step_012	data/hot3d/clip-001899/step_012.jpg	[ "{\"from\":\"human\",\"value\":\"<image>\\nWhat action should the robot take to `put the glasses on the table`\"}", "{\"from\":\"gpt\",\"raw_actions\":[0.011491724,-0.0030624971,-0.0175770884,-0.0208063142,-0.0415289088,-0.0150623195]}" ]
data/hot3d/clip-001899/step_013	data/hot3d/clip-001899/step_013.jpg	[ "{\"from\":\"human\",\"value\":\"<image>\\nWhat action should the robot take to `put the glasses on the table`\"}", "{\"from\":\"gpt\",\"raw_actions\":[-0.01160538,-0.0191560005,0.0928889725,0.037612709,0.2091736451,0.06824786]}" ]
data/hot3d/clip-001899/step_014	data/hot3d/clip-001899/step_014.jpg	[ "{\"from\":\"human\",\"value\":\"<image>\\nWhat action should the robot take to `put the glasses on the table`\"}", "{\"from\":\"gpt\",\"raw_actions\":[0.0024815351,-0.0161271053,0.0905276778,0.0303301361,0.2362230612,0.0873989349]}" ]
data/hot3d/clip-001899/step_015	data/hot3d/clip-001899/step_015.jpg	[ "{\"from\":\"human\",\"value\":\"<image>\\nWhat action should the robot take to `put the glasses on the table`\"}", "{\"from\":\"gpt\",\"raw_actions\":[0.0084830678,-0.0138300721,-0.0048633937,-0.0108305146,0.088093244,0.0733250707]}" ]
data/hot3d/clip-001899/step_016	data/hot3d/clip-001899/step_016.jpg	[ "{\"from\":\"human\",\"value\":\"<image>\\nWhat action should the robot take to `put the glasses on the table`\"}", "{\"from\":\"gpt\",\"raw_actions\":[-0.0050653572,-0.029469216,-0.0698215863,0.0837620173,-0.0082198802,-0.0476347306]}" ]
data/hot3d/clip-001899/step_017	data/hot3d/clip-001899/step_017.jpg	[ "{\"from\":\"human\",\"value\":\"<image>\\nWhat action should the robot take to `put the glasses on the table`\"}", "{\"from\":\"gpt\",\"raw_actions\":[-0.0058799447,-0.0428030727,0.0076429644,0.1649616707,0.0338679323,-0.0485134726]}" ]
data/hot3d/clip-001899/step_018	data/hot3d/clip-001899/step_018.jpg	[ "{\"from\":\"human\",\"value\":\"<image>\\nWhat action should the robot take to `put the glasses on the table`\"}", "{\"from\":\"gpt\",\"raw_actions\":[0.008940967,-0.0473375039,0.0588100098,0.078241614,0.2713492916,-0.0949923579]}" ]
data/hot3d/clip-001899/step_019	data/hot3d/clip-001899/step_019.jpg	[ "{\"from\":\"human\",\"value\":\"<image>\\nWhat action should the robot take to `put the glasses on the table`\"}", "{\"from\":\"gpt\",\"raw_actions\":[0.0115237923,-0.0207978129,0.0631525961,0.184892687,0.1273620429,-0.1972924278]}" ]
data/hot3d/clip-001899/step_020	data/hot3d/clip-001899/step_020.jpg	[ "{\"from\":\"human\",\"value\":\"<image>\\nWhat action should the robot take to `put the glasses on the table`\"}", "{\"from\":\"gpt\",\"raw_actions\":[0.0144210018,-0.003205427,0.0283398308,0.0807822779,0.1289956683,-0.1426303786]}" ]
data/hot3d/clip-001899/step_021	data/hot3d/clip-001899/step_021.jpg	[ "{\"from\":\"human\",\"value\":\"<image>\\nWhat action should the robot take to `put the glasses on the table`\"}", "{\"from\":\"gpt\",\"raw_actions\":[-0.0016032741,0.0058415015,-0.013687467,0.0480003317,0.2078018081,0.0245425846]}" ]
data/hot3d/clip-001899/step_022	data/hot3d/clip-001899/step_022.jpg	[ "{\"from\":\"human\",\"value\":\"<image>\\nWhat action should the robot take to `put the glasses on the table`\"}", "{\"from\":\"gpt\",\"raw_actions\":[-0.0688951779,0.0701702264,-0.0083054236,-0.0178384132,-0.0910934686,0.4114873047]}" ]
data/hot3d/clip-001905/step_000	data/hot3d/clip-001905/step_000.jpg	[ "{\"from\":\"human\",\"value\":\"<image>\\nWhat action should the robot take to `pick up the flask`\"}", "{\"from\":\"gpt\",\"raw_actions\":[-0.0000065977,0.0400718282,0.0156792954,0.0717485565,-0.0910943456,0.1506546905]}" ]
data/hot3d/clip-001905/step_001	data/hot3d/clip-001905/step_001.jpg	[ "{\"from\":\"human\",\"value\":\"<image>\\nWhat action should the robot take to `pick up the flask`\"}", "{\"from\":\"gpt\",\"raw_actions\":[0.0238251739,0.0159490506,-0.0531146203,0.2741933773,-0.0525677965,0.1748372084]}" ]
data/hot3d/clip-001905/step_002	data/hot3d/clip-001905/step_002.jpg	[ "{\"from\":\"human\",\"value\":\"<image>\\nWhat action should the robot take to `pick up the flask`\"}", "{\"from\":\"gpt\",\"raw_actions\":[0.0607516053,0.0099715815,-0.0986943085,0.3496130848,-0.0587761322,0.2330424504]}" ]
data/hot3d/clip-001905/step_003	data/hot3d/clip-001905/step_003.jpg	[ "{\"from\":\"human\",\"value\":\"<image>\\nWhat action should the robot take to `pick up the flask`\"}", "{\"from\":\"gpt\",\"raw_actions\":[0.0678002926,0.0219005749,-0.0992388112,0.3519732626,-0.0057657035,0.2210900524]}" ]
data/hot3d/clip-001905/step_004	data/hot3d/clip-001905/step_004.jpg	[ "{\"from\":\"human\",\"value\":\"<image>\\nWhat action should the robot take to `pick up the flask`\"}", "{\"from\":\"gpt\",\"raw_actions\":[0.037549084,0.053024583,-0.0589070009,0.3406976743,0.0187316092,0.1851300616]}" ]
data/hot3d/clip-001905/step_005	data/hot3d/clip-001905/step_005.jpg	[ "{\"from\":\"human\",\"value\":\"<image>\\nWhat action should the robot take to `pick up the flask`\"}", "{\"from\":\"gpt\",\"raw_actions\":[0.0056739077,0.0762836722,-0.0100150435,0.2935641715,-0.012299917,0.201574617]}" ]
data/hot3d/clip-001905/step_006	data/hot3d/clip-001905/step_006.jpg	[ "{\"from\":\"human\",\"value\":\"<image>\\nWhat action should the robot take to `pick up the flask`\"}", "{\"from\":\"gpt\",\"raw_actions\":[-0.0208764144,0.100176289,0.0322665849,0.2877707863,-0.0477019744,0.2313270611]}" ]
data/hot3d/clip-001905/step_007	data/hot3d/clip-001905/step_007.jpg	[ "{\"from\":\"human\",\"value\":\"<image>\\nWhat action should the robot take to `pick up the flask`\"}", "{\"from\":\"gpt\",\"raw_actions\":[-0.0357340497,0.1157589373,0.053870008,0.2485777006,-0.0815717706,0.1748338352]}" ]
data/hot3d/clip-001905/step_008	data/hot3d/clip-001905/step_008.jpg	[ "{\"from\":\"human\",\"value\":\"<image>\\nWhat action should the robot take to `pick up the flask`\"}", "{\"from\":\"gpt\",\"raw_actions\":[-0.0414605238,0.102740155,0.0335040739,0.1681290084,-0.025370723,0.1820211485]}" ]
data/hot3d/clip-001905/step_009	data/hot3d/clip-001905/step_009.jpg	[ "{\"from\":\"human\",\"value\":\"<image>\\nWhat action should the robot take to `pick up the flask`\"}", "{\"from\":\"gpt\",\"raw_actions\":[-0.0401002294,0.0717699547,-0.002436408,-0.0079781725,-0.0590667693,0.2574336942]}" ]
data/hot3d/clip-001905/step_010	data/hot3d/clip-001905/step_010.jpg	[ "{\"from\":\"human\",\"value\":\"<image>\\nWhat action should the robot take to `pick up the flask`\"}", "{\"from\":\"gpt\",\"raw_actions\":[-0.014532119,0.027916226,-0.0229842415,-0.0655842269,-0.2314340093,0.1895321843]}" ]
data/hot3d/clip-001905/step_011	data/hot3d/clip-001905/step_011.jpg	[ "{\"from\":\"human\",\"value\":\"<image>\\nWhat action should the robot take to `pick up the flask`\"}", "{\"from\":\"gpt\",\"raw_actions\":[0.0063064346,0.0038894383,-0.0342337291,0.0619596968,-0.2177622291,0.123770893]}" ]
data/hot3d/clip-001905/step_012	data/hot3d/clip-001905/step_012.jpg	[ "{\"from\":\"human\",\"value\":\"<image>\\nWhat action should the robot take to `pick up the flask`\"}", "{\"from\":\"gpt\",\"raw_actions\":[0.0096135492,0.0049329859,-0.0367700676,0.0779368016,-0.2064620201,0.0536760722]}" ]
data/hot3d/clip-001905/step_013	data/hot3d/clip-001905/step_013.jpg	[ "{\"from\":\"human\",\"value\":\"<image>\\nWhat action should the robot take to `pick up the flask`\"}", "{\"from\":\"gpt\",\"raw_actions\":[0.0118329461,0.002677527,-0.0325691498,0.0370286492,-0.2422158943,0.0662176897]}" ]
data/hot3d/clip-001905/step_014	data/hot3d/clip-001905/step_014.jpg	[ "{\"from\":\"human\",\"value\":\"<image>\\nWhat action should the robot take to `pick up the flask`\"}", "{\"from\":\"gpt\",\"raw_actions\":[0.0141139665,0.0052852639,-0.0107333543,0.0336398866,-0.1546502845,0.0487027646]}" ]
data/hot3d/clip-001905/step_015	data/hot3d/clip-001905/step_015.jpg	[ "{\"from\":\"human\",\"value\":\"<image>\\nWhat action should the robot take to `pick up the flask`\"}", "{\"from\":\"gpt\",\"raw_actions\":[-0.0021637265,-0.0026937619,0.004260899,-0.0058762138,0.0368808402,0.0142567752]}" ]
data/hot3d/clip-001905/step_016	data/hot3d/clip-001905/step_016.jpg	[ "{\"from\":\"human\",\"value\":\"<image>\\nWhat action should the robot take to `pick up the flask`\"}", "{\"from\":\"gpt\",\"raw_actions\":[-0.0110206552,-0.0107795295,-0.0067383176,-0.0063835636,0.1311403148,-0.0181058317]}" ]