Egocentric Manipulation Sample

Public   9 clips  ·  169 files  ·  19.4 GB  ·  2026-06-23

Evaluation sample for robotics and embodied AI. Monocular smartphone ego video from real Indian workplaces.

aws s3 sync s3://ggn-egocentric-data-sample/sample_data_june ./Master_Sample_v1 --no-sign-request

Quick start

Clips 01–09

#TaskEnvironmentDeviceSessionVideoOverlayBoxesMetadata
01Shuttle tube packagingfactorySamsung Galaxy S24GGN_20260618_S01videooverlayboxesmetadata
02Industrial sewingfactorySamsung Galaxy S24GGN_20260618_S02videooverlayboxesmetadata
03Heat gun & batchingfactorySamsung Galaxy S24GGN_20260618_S03videooverlayboxesmetadata
04Garment ironing & packingfactorySamsung Galaxy S24GGN_20260618_S04videooverlayboxesmetadata
05Commercial cateringrestaurantiPhone 16 Pro MaxGGN_20260620_S05videooverlayboxesmetadata
06Cane weavingroadside shopiPhone 16 Pro MaxGGN_20260620_S06videooverlayboxesmetadata
07Car detailingcar showroomSamsung Galaxy S24GGN_20260620_S07videooverlayboxesmetadata
08Primer & paintingrepair shopSamsung Galaxy S24GGN_20260620_S08videooverlayboxesmetadata
09Denting & filingroadside repairSamsung Galaxy S24GGN_20260620_S09videooverlayboxesmetadata

Complete file tree

PathSizeTypeDescription
📁 annotations/ — All temporal and spatial labels (JSONL, CSV, summaries)
📁 timestamps/ — Per-clip frame_idx ↔ timestamp_sec CSVs (clip-relative time)
📄 sample_01_shuttle_tube_packaging_timestamps.csv274.2 KBCSVFrame index and timestamp (seconds) mapping
📄 sample_02_industrial_sewing_machine_timestamps.csv260.6 KBCSVFrame index and timestamp (seconds) mapping
📄 sample_03_heatgun_and_batching_timestamps.csv215.6 KBCSVFrame index and timestamp (seconds) mapping
📄 sample_04_garment_ironing_and_packing_timestamps.csv420.9 KBCSVFrame index and timestamp (seconds) mapping
📄 sample_05_commercial_catering_timestamps.csv294.0 KBCSVFrame index and timestamp (seconds) mapping
📄 sample_06_cane_weaving_timestamps.csv271.4 KBCSVFrame index and timestamp (seconds) mapping
📄 sample_07_car_detailing_timestamps.csv207.7 KBCSVFrame index and timestamp (seconds) mapping
📄 sample_08_primer_and_painting_timestamps.csv227.2 KBCSVFrame index and timestamp (seconds) mapping
📄 sample_09_denting_and_filing_timestamps.csv227.6 KBCSVFrame index and timestamp (seconds) mapping
📄 action_segments.jsonl49.7 KBJSONLTemporal verb–noun manipulation phases (all clips)
📄 captions.jsonl3.1 KBJSONLOne natural-language paragraph per clip
📄 hand_boxes_summary.json3.4 KBJSON
📄 hand_keypoints_summary.json3.8 KBJSONRoll-up — hands detected %, two-hands % per clip
📄 object_boxes_summary.json3.4 KBJSONRoll-up — detection counts per clip and class
📄 sample_01_shuttle_tube_packaging_hand_boxes.jsonl3.1 MBJSONL2D hand bounding boxes (left/right, per frame, from keypoints)
📄 sample_01_shuttle_tube_packaging_hand_keypoints.jsonl13.4 MBJSONL2D hand keypoints — 21 landmarks per hand, one row per frame
📄 sample_01_shuttle_tube_packaging_hand_object_contact.jsonl22.0 KBJSONLHand–object contact samples (derived from bbox overlap)
📄 sample_01_shuttle_tube_packaging_object_boxes.jsonl1.5 MBJSONL2D object bounding boxes + track IDs (auto-generated, sampled)
📄 sample_02_industrial_sewing_machine_hand_boxes.jsonl1.7 MBJSONL2D hand bounding boxes (left/right, per frame, from keypoints)
📄 sample_02_industrial_sewing_machine_hand_keypoints.jsonl8.1 MBJSONL2D hand keypoints — 21 landmarks per hand, one row per frame
📄 sample_02_industrial_sewing_machine_hand_object_contact.jsonl14.0 KBJSONLHand–object contact samples (derived from bbox overlap)
📄 sample_02_industrial_sewing_machine_object_boxes.jsonl623.2 KBJSONL2D object bounding boxes + track IDs (auto-generated, sampled)
📄 sample_03_heatgun_and_batching_hand_boxes.jsonl790.6 KBJSONL2D hand bounding boxes (left/right, per frame, from keypoints)
📄 sample_03_heatgun_and_batching_hand_keypoints.jsonl4.5 MBJSONL2D hand keypoints — 21 landmarks per hand, one row per frame
📄 sample_03_heatgun_and_batching_hand_object_contact.jsonl3.2 KBJSONLHand–object contact samples (derived from bbox overlap)
📄 sample_03_heatgun_and_batching_object_boxes.jsonl92.3 KBJSONL2D object bounding boxes + track IDs (auto-generated, sampled)
📄 sample_04_garment_ironing_and_packing_hand_boxes.jsonl1.6 MBJSONL2D hand bounding boxes (left/right, per frame, from keypoints)
📄 sample_04_garment_ironing_and_packing_hand_keypoints.jsonl8.7 MBJSONL2D hand keypoints — 21 landmarks per hand, one row per frame
📄 sample_04_garment_ironing_and_packing_hand_object_contact.jsonl5.5 KBJSONLHand–object contact samples (derived from bbox overlap)
📄 sample_04_garment_ironing_and_packing_object_boxes.jsonl169.2 KBJSONL2D object bounding boxes + track IDs (auto-generated, sampled)
📄 sample_05_commercial_catering_hand_boxes.jsonl873.5 KBJSONL2D hand bounding boxes (left/right, per frame, from keypoints)
📄 sample_05_commercial_catering_hand_keypoints.jsonl5.3 MBJSONL2D hand keypoints — 21 landmarks per hand, one row per frame
📄 sample_05_commercial_catering_hand_object_contact.jsonl686 BJSONLHand–object contact samples (derived from bbox overlap)
📄 sample_05_commercial_catering_object_boxes.jsonl80.0 KBJSONL2D object bounding boxes + track IDs (auto-generated, sampled)
📄 sample_06_cane_weaving_hand_boxes.jsonl2.2 MBJSONL2D hand bounding boxes (left/right, per frame, from keypoints)
📄 sample_06_cane_weaving_hand_keypoints.jsonl10.5 MBJSONL2D hand keypoints — 21 landmarks per hand, one row per frame
📄 sample_06_cane_weaving_hand_object_contact.jsonl10.3 KBJSONLHand–object contact samples (derived from bbox overlap)
📄 sample_06_cane_weaving_object_boxes.jsonl294.4 KBJSONL2D object bounding boxes + track IDs (auto-generated, sampled)
📄 sample_07_car_detailing_hand_boxes.jsonl1.8 MBJSONL2D hand bounding boxes (left/right, per frame, from keypoints)
📄 sample_07_car_detailing_hand_keypoints.jsonl8.4 MBJSONL2D hand keypoints — 21 landmarks per hand, one row per frame
📄 sample_07_car_detailing_hand_object_contact.jsonl36.6 KBJSONLHand–object contact samples (derived from bbox overlap)
📄 sample_07_car_detailing_object_boxes.jsonl123.3 KBJSONL2D object bounding boxes + track IDs (auto-generated, sampled)
📄 sample_08_primer_and_painting_hand_boxes.jsonl1.4 MBJSONL2D hand bounding boxes (left/right, per frame, from keypoints)
📄 sample_08_primer_and_painting_hand_keypoints.jsonl7.1 MBJSONL2D hand keypoints — 21 landmarks per hand, one row per frame
📄 sample_08_primer_and_painting_hand_object_contact.jsonl17.0 KBJSONLHand–object contact samples (derived from bbox overlap)
📄 sample_08_primer_and_painting_object_boxes.jsonl315.8 KBJSONL2D object bounding boxes + track IDs (auto-generated, sampled)
📄 sample_09_denting_and_filing_hand_boxes.jsonl736.3 KBJSONL2D hand bounding boxes (left/right, per frame, from keypoints)
📄 sample_09_denting_and_filing_hand_keypoints.jsonl4.4 MBJSONL2D hand keypoints — 21 landmarks per hand, one row per frame
📄 sample_09_denting_and_filing_hand_object_contact.jsonl2.8 KBJSONLHand–object contact samples (derived from bbox overlap)
📄 sample_09_denting_and_filing_object_boxes.jsonl139.9 KBJSONL2D object bounding boxes + track IDs (auto-generated, sampled)
📄 validation_report.json5.9 KBJSON
📁 consent/ — Commercial AI training consent summary
📄 worker_consent_summary.md1.4 KBMDWorker consent for commercial AI training
📁 metadata/ — Per-clip JSON — device, QA flags, session, manipulator stats, SHA256
📄 sample_01_shuttle_tube_packaging.json3.0 KBJSONPer-clip metadata — QA, session, integrity, manipulator stats
📄 sample_02_industrial_sewing_machine.json3.1 KBJSONPer-clip metadata — QA, session, integrity, manipulator stats
📄 sample_03_heatgun_and_batching.json3.1 KBJSONPer-clip metadata — QA, session, integrity, manipulator stats
📄 sample_04_garment_ironing_and_packing.json3.0 KBJSONPer-clip metadata — QA, session, integrity, manipulator stats
📄 sample_05_commercial_catering.json3.1 KBJSONPer-clip metadata — QA, session, integrity, manipulator stats
📄 sample_06_cane_weaving.json3.1 KBJSONPer-clip metadata — QA, session, integrity, manipulator stats
📄 sample_07_car_detailing.json3.1 KBJSONPer-clip metadata — QA, session, integrity, manipulator stats
📄 sample_08_primer_and_painting.json3.1 KBJSONPer-clip metadata — QA, session, integrity, manipulator stats
📄 sample_09_denting_and_filing.json3.1 KBJSONPer-clip metadata — QA, session, integrity, manipulator stats
📁 overlays/ — Hand skeleton preview MP4s — 2D landmarks visualized on video
📄 sample_01_shuttle_tube_packaging_overlay.mp4539.1 MBMP4Hand skeleton overlay preview video
📄 sample_02_industrial_sewing_machine_overlay.mp4278.4 MBMP4Hand skeleton overlay preview video
📄 sample_03_heatgun_and_batching_overlay.mp4480.8 MBMP4Hand skeleton overlay preview video
📄 sample_04_garment_ironing_and_packing_overlay.mp4326.9 MBMP4Hand skeleton overlay preview video
📄 sample_05_commercial_catering_overlay.mp4363.7 MBMP4Hand skeleton overlay preview video
📄 sample_06_cane_weaving_overlay.mp4394.6 MBMP4Hand skeleton overlay preview video
📄 sample_07_car_detailing_overlay.mp4777.4 MBMP4Hand skeleton overlay preview video
📄 sample_08_primer_and_painting_overlay.mp4887.3 MBMP4Hand skeleton overlay preview video
📄 sample_09_denting_and_filing_overlay.mp4967.6 MBMP4Hand skeleton overlay preview video
📁 previews/ — Object bounding-box overlay preview MP4s (full-length, per clip)
📄 sample_01_shuttle_tube_packaging_boxes_preview.mp4533.8 MBMP4Overlay preview video — object + hand boxes (full length, normal speed)
📄 sample_02_industrial_sewing_machine_boxes_preview.mp4273.4 MBMP4Overlay preview video — object + hand boxes (full length, normal speed)
📄 sample_03_heatgun_and_batching_boxes_preview.mp4479.4 MBMP4Overlay preview video — object + hand boxes (full length, normal speed)
📄 sample_04_garment_ironing_and_packing_boxes_preview.mp4318.0 MBMP4Overlay preview video — object + hand boxes (full length, normal speed)
📄 sample_05_commercial_catering_boxes_preview.mp4364.0 MBMP4Overlay preview video — object + hand boxes (full length, normal speed)
📄 sample_06_cane_weaving_boxes_preview.mp4390.7 MBMP4Overlay preview video — object + hand boxes (full length, normal speed)
📄 sample_07_car_detailing_boxes_preview.mp4762.1 MBMP4Overlay preview video — object + hand boxes (full length, normal speed)
📄 sample_08_primer_and_painting_boxes_preview.mp4883.9 MBMP4Overlay preview video — object + hand boxes (full length, normal speed)
📄 sample_09_denting_and_filing_boxes_preview.mp4964.8 MBMP4Overlay preview video — object + hand boxes (full length, normal speed)
📁 qa/ — Structural QA roll-up — resolution, duration alignment, CI pass flags
📄 qa_report.json6.2 KBJSONMachine-readable QA table
📄 qa_report.md1.3 KBMDHuman-readable QA summary
📁 review_frames/ — Segment snapshots for visual QA (clips 03, 06)
📁 sample_03_heatgun_and_batching/ — Review frames — heat gun and shuttle tube clip
📄 seg001_heat_gun_t5.5s.jpg469.8 KBJPGSegment snapshot for visual QA
📄 seg002_shuttle_tube_t15.0s.jpg563.0 KBJPGSegment snapshot for visual QA
📄 seg003_heat_gun_t22.0s.jpg536.1 KBJPGSegment snapshot for visual QA
📄 seg004_heat_gun_t30.5s.jpg601.2 KBJPGSegment snapshot for visual QA
📄 seg005_shuttle_tube_t42.0s.jpg509.2 KBJPGSegment snapshot for visual QA
📄 seg006_heat_gun_t54.5s.jpg579.8 KBJPGSegment snapshot for visual QA
📄 seg007_heat_gun_t67.0s.jpg536.2 KBJPGSegment snapshot for visual QA
📄 seg008_heat_gun_t81.0s.jpg608.8 KBJPGSegment snapshot for visual QA
📄 seg009_heat_gun_t94.5s.jpg547.3 KBJPGSegment snapshot for visual QA
📄 seg010_shuttle_tube_t109.0s.jpg455.6 KBJPGSegment snapshot for visual QA
📄 seg011_heat_gun_t126.0s.jpg542.6 KBJPGSegment snapshot for visual QA
📄 seg012_heat_gun_t141.5s.jpg523.2 KBJPGSegment snapshot for visual QA
📄 seg013_heat_gun_t158.0s.jpg477.6 KBJPGSegment snapshot for visual QA
📄 seg014_heat_gun_t174.5s.jpg537.5 KBJPGSegment snapshot for visual QA
📄 seg015_heat_gun_t193.5s.jpg460.6 KBJPGSegment snapshot for visual QA
📄 seg016_heat_gun_t211.5s.jpg489.5 KBJPGSegment snapshot for visual QA
📄 seg017_rope_t225.0s.jpg453.4 KBJPGSegment snapshot for visual QA
📄 seg018_shuttle_tube_t236.5s.jpg468.5 KBJPGSegment snapshot for visual QA
📄 seg019_shuttle_tube_t248.0s.jpg554.9 KBJPGSegment snapshot for visual QA
📄 seg020_shuttle_tube_t260.0s.jpg481.9 KBJPGSegment snapshot for visual QA
📄 seg021_shuttle_tube_t273.0s.jpg391.3 KBJPGSegment snapshot for visual QA
📄 seg022_shuttle_tube_t286.0s.jpg444.4 KBJPGSegment snapshot for visual QA
📄 seg023_shuttle_tube_t294.5s.jpg452.1 KBJPGSegment snapshot for visual QA
📄 seg024_shuttle_tube_t302.5s.jpg515.7 KBJPGSegment snapshot for visual QA
📄 segment_guide.json9.7 KBJSONSegment to object-class reference for QA
📁 sample_06_cane_weaving/ — Review frames — cane weaving clip
📄 seg001_hammer_t3.5s.jpg572.3 KBJPGSegment snapshot for visual QA
📄 seg002_chair_t8.0s.jpg653.1 KBJPGSegment snapshot for visual QA
📄 seg003_nail_t12.5s.jpg583.8 KBJPGSegment snapshot for visual QA
📄 seg004_wooden_strap_t18.5s.jpg563.2 KBJPGSegment snapshot for visual QA
📄 seg005_wooden_strap_t24.0s.jpg577.9 KBJPGSegment snapshot for visual QA
📄 seg006_nail_t32.0s.jpg684.1 KBJPGSegment snapshot for visual QA
📄 seg007_nail_t42.0s.jpg564.1 KBJPGSegment snapshot for visual QA
📄 seg008_wooden_strap_t53.0s.jpg604.1 KBJPGSegment snapshot for visual QA
📄 seg009_nail_t67.5s.jpg599.9 KBJPGSegment snapshot for visual QA
📄 seg010_chair_t78.5s.jpg487.1 KBJPGSegment snapshot for visual QA
📄 seg011_nail_t86.0s.jpg559.2 KBJPGSegment snapshot for visual QA
📄 seg012_nail_t97.0s.jpg579.9 KBJPGSegment snapshot for visual QA
📄 seg013_blade_t107.0s.jpg439.8 KBJPGSegment snapshot for visual QA
📄 seg014_nail_t115.5s.jpg577.2 KBJPGSegment snapshot for visual QA
📄 seg015_wooden_strap_t126.0s.jpg656.8 KBJPGSegment snapshot for visual QA
📄 seg016_cutter_t144.5s.jpg632.8 KBJPGSegment snapshot for visual QA
📄 seg017_chair_t163.0s.jpg617.5 KBJPGSegment snapshot for visual QA
📄 seg018_nail_t183.0s.jpg486.8 KBJPGSegment snapshot for visual QA
📄 seg019_nail_t201.5s.jpg529.4 KBJPGSegment snapshot for visual QA
📄 seg020_cutter_t210.0s.jpg580.1 KBJPGSegment snapshot for visual QA
📄 seg021_nail_t221.5s.jpg612.1 KBJPGSegment snapshot for visual QA
📄 seg022_nail_t248.5s.jpg468.9 KBJPGSegment snapshot for visual QA
📄 seg023_wooden_strap_t275.5s.jpg510.6 KBJPGSegment snapshot for visual QA
📄 seg024_nail_t302.0s.jpg516.5 KBJPGSegment snapshot for visual QA
📄 seg025_nail_t332.9s.jpg622.3 KBJPGSegment snapshot for visual QA
📄 segment_guide.json8.8 KBJSONSegment to object-class reference for QA
📁 schema/ — Field definitions for ingest and audit
📄 action_taxonomy.md1.0 KBMDSchema and field definitions
📄 annotation_schema.md956 BMDSchema and field definitions
📄 capture_qa_schema.md1.8 KBMDSchema and field definitions
📄 delivery_schema.md518 BMDSchema and field definitions
📄 hand_boxes_schema.md1.8 KBMDSchema and field definitions
📄 object_boxes_schema.md2.5 KBMDSchema and field definitions
📄 object_taxonomy.md1.2 KBMDSchema and field definitions
📄 session_schema.md1.4 KBMDSchema and field definitions
📁 sessions/ — Session-layer JSON — capture date, lighting, collector, mount
📄 GGN_20260618_S01.json1.0 KBJSONSession capture metadata
📄 GGN_20260618_S02.json1.0 KBJSONSession capture metadata
📄 GGN_20260618_S03.json1.0 KBJSONSession capture metadata
📄 GGN_20260618_S04.json1.0 KBJSONSession capture metadata
📄 GGN_20260620_S05.json1.0 KBJSONSession capture metadata
📄 GGN_20260620_S06.json1009 BJSONSession capture metadata
📄 GGN_20260620_S07.json1009 BJSONSession capture metadata
📄 GGN_20260620_S08.json1.0 KBJSONSession capture metadata
📄 GGN_20260620_S09.json1.0 KBJSONSession capture metadata
📄 session_index.json10.0 KBJSONIndex of all 9 capture sessions
📁 videos/ — Final-cut ego MP4s (H.264, no audio) — 9 workplace manipulation clips
📁 review/ — 3 items
📄 sample_04_cfr_preserve.mp4540.5 MBMP4Deliverable ego video (no audio)
📄 sample_04_cfr_test.mp4465.0 MBMP4Deliverable ego video (no audio)
📄 sample_04_vfr_original.mp4714.6 MBMP4Deliverable ego video (no audio)
📄 sample_01_shuttle_tube_packaging.mp4740.5 MBMP4Deliverable ego video (no audio)
📄 sample_02_industrial_sewing_machine.mp4507.3 MBMP4Deliverable ego video (no audio)
📄 sample_03_heatgun_and_batching.mp4629.0 MBMP4Deliverable ego video (no audio)
📄 sample_04_garment_ironing_and_packing.mp4540.5 MBMP4Deliverable ego video (no audio)
📄 sample_05_commercial_catering.mp4756.4 MBMP4Deliverable ego video (no audio)
📄 sample_06_cane_weaving.mp4384.9 MBMP4Deliverable ego video (no audio)
📄 sample_07_car_detailing.mp41.6 GBMP4Deliverable ego video (no audio)
📄 sample_08_primer_and_painting.mp41.4 GBMP4Deliverable ego video (no audio)
📄 sample_09_denting_and_filing.mp41.4 GBMP4Deliverable ego video (no audio)
📄 camera_config.json2.9 KBJSONMonocular rig config — device, resolution, fps, mount per clip
📄 DATACARD.md4.3 KBMD
📄 delivery_manifest.json372 BJSONPack header — version, clip count, license, integrity hash
📄 DELIVERY_OVERVIEW.md4.3 KBMDPipeline, package layout, QA flag definitions
📄 manifest.csv6.2 KBCSVTabular provenance export
📄 manifest.json13.9 KBJSONProvenance per clip — source file, cut times, SHA256
📄 README.md1.7 KBMDClip list, device map, layer summary

Data layers

LayerLocationDescription
Source videosvideos/9 MP4s, H.264, audio stripped
Hand overlaysoverlays/2D hand skeleton preview videos
Box overlayspreviews/Full-length videos with object + hand boxes drawn on each clip (normal speed)
Action segmentsannotations/action_segments.jsonl218 verb–noun manipulation phases
Captionsannotations/captions.jsonlOne paragraph per clip
Hand keypointsannotations/*_hand_keypoints.jsonl21 landmarks per hand, per frame
Object boxesannotations/*_object_boxes.jsonl2D object bbox + track IDs (sampled)
Hand boxesannotations/*_hand_boxes.jsonl2D left/right hand bbox, per frame (from keypoints)
Hand–object contactannotations/*_hand_object_contact.jsonlDerived contact samples
Timestampsannotations/timestamps/Frame ↔ second CSVs
Metadatametadata/QA flags, session, SHA256, manipulator stats
Sessionssessions/Capture date, lighting, collector, mount