Industrial First-Person Video • 6,000 to 8,000 Hrs/Day Throughput
Proven across leading Vision-Language-Action (VLA) foundation models, Video-LLaVA, and spatial robotics policies. Backed by 30+ accredited field agencies & 90,000+ consented operatives across India.
🧵 Garments, Textiles & Power Loom Weaving
Industrial sewing machine stitching, overlocking, cloth cutting, pattern grading, loom operation, steam pressing, and final packaging. Captures non-rigid material contact dynamics and hand dexterity.
Ongoing capture campaigns & egocentric datasets
A real-time snapshot of active first-person (egocentric) POV and robotics training data collection drives executed across India.
State-of-the-Art 2,500 Sq Ft AI Lab & Embodied Robotics Studio
Blue Projects is expanding with a flagship 2,500 sq ft indoor testing studio featuring 5 specialized staging zones: domestic kitchen sets, teleoperation robotic workbenches, 19-point IMU mocap volumes, GPU control centers, and a 50-station data annotation hub.
India-Wide Data Acquisition Network
We have established a linked network of trained participants, workspace locations, and small enterprises across diverse regions.
Unorganized Sector Workers
Consented and linked operatives for manual agricultural, construction, retail, and domestic scenario capture drives.
Organized Sector Workers
Access to structured commercial environments including textile mills, manufacturing lines, and warehouse personnel.
MSME Partners
Registered micro, small, and medium enterprises cooperating for industrial task recordings and spatial scanning.
Industries Linked
Partnership channels with commercial factories, processing units, and specialized manufacturing plants.
Calibrated Capture Hardware
Our standardized sensor hardware kits deployed in the field to guarantee time-synchronized, distortion-free dataset feeds.
Headmounted camera
High-definition multi-lens egocentric headsets capturing first-person POV video sequences and spatial depth maps.
Mobile-based capture
Smartphone headstrap rigs designed to capture first-person visual perspectives directly through mobile capture clients.
GoPro headmount camera
Rugged wide-angle action cameras mounted on headbands for high-activity indoor/outdoor human POV logging.
Mobile app robotics capture
Handheld camera and telemetry logging apps mapping bimanual trajectories and orientations for robotic training.
Engineered Dataset Samples for Foundation Models
Explore sample downloads, sensor schemas, and video streams across robotics teleoperation, spatial computing, egocentric vision, and Indic speech.
Looking for Custom Sensor Configurations?
Tell us your target joint angles, frame rates, or dialect matrices and access our live Google Drive repository.
Targeted acquisition in complex real-world environments
No synthetic data, no web-scraping — every dataset is captured by trained field operatives from real consented participants under rigorous QA audits.
👁️ Egocentric / First-Person POV
Chest rigs, head mounts, and eyewear capturing commutes, manual labor, and domestic tasks—the exact visual coordinates your robotics models require to learn.
🚶 Human Behaviour Video
Subjects walking, cooking, shopping, driving, and interacting across urban, semi-urban, and rural Indian environments. Visual complexity at scale.
🤖 Robotics & Manipulation Data
Fine-grained hand-object interactions, bimanual sorting, and tool usage across textile units, agricultural fields, and warehousing operations.
🏷️ Annotation & Labelling
Skeletal keypoint landmarks, bounding boxes, polygon segmentation, temporal boundary tags, and frame metadata mapping.
😊 Activity & Gesture Vocabulary
Multi-camera capturing of gestures, micro-expressions, and physical trajectories across diverse age, gender, and regional demographics.
🎙️ Speech & Dialect Audio
Acoustic data recorded in natural settings (home, street, work) across 10 major Indian languages to close the ASR accent gap.
Be the first to explore
Humanoid Bimanual Teleoperation
High-frequency 50Hz dual-arm leader-follower teleoperation trajectory logging for dexterous humanoid hands.
3D LiDAR Point Cloud & Cuboids
Calibrated 32-channel 3D point cloud scans with 3D bounding box annotations and sensor camera fusion.
Egocentric First-Person POV Video
4K 60fps first-person perspective video captured from head-mounted EgoCAMs across 8 Indian states for VLA models.
Indic Multilingual Audio & IPA
Acoustic speech corpora recorded across 22 Indic scheduled languages with IPA phonetic alignments.
CV & Data Annotation Services
Bounding box, polygon segmentation, 3D LiDAR cuboids, and 21-keypoint skeletal hand pose tracking.
Human-in-the-Loop & RLHF Studio
Pairwise preference ranking, DPO tuples, process supervision, and red-teaming safety evaluations.
Process Supervision & CoT
Step-by-step reasoning verification, math/logic CoT traces, and PRM process reward models.
Procedural Synthetic Data
3D procedural scene synthesis, domain randomization, and synthetic edge-case physical injection.
Robotic Teleoperation Pipeline
Bilateral leader-follower rigs, 6-axis force-torque load cells, and low-latency haptic teleoperation control.
Digital Twins & Physics Simulators
High-rate 1kHz physics step trajectories, URDF/MJCF kinematic articulation, and tactile contact solvers.
Domain-Expert Legal & Tax RLHF
Bar-certified lawyers & Chartered Accountants verifying statutory citations, GST reasoning, and contract logic.
Localized Dialect Audio
Natural multi-speaker conversational recordings capturing vernacular codeswitching and environmental noise.
Human Behaviour & Gestures
Full-body motion kinetics, gait analysis, and physical task taxonomy across real-world Indian work settings.
Surgical & Medical Vision
Physician-annotated laparoscopic video, anatomical segmentation, and robotic instrument tracking datasets.
Synthetic-to-Real Sim2Real
Domain gap reduction, neural rendering texture randomization, and real-world transfer validation.
3D Output Formats & Schemas
Standardized dataset specs and converters for HDF5, OpenUSD, Parquet, ROSbag, and Point Cloud formats.