ATTRSENSE
EgOmini 4X

EgOmini 4X

A multi-camera data capture device for embodied AI and robot learning. Four global-shutter cameras in a ring layout, synchronized to the microsecond, reliably capture high-quality, frame-aligned egocentric multi-view video, providing a dependable data foundation for imitation learning, visual navigation, and world model training.

PRODUCT INTRODUCTION

EgOmini 4X

EgOmini 4XMulti-Camera Embodied Data Capture Device

With four global-shutter cameras and microsecond-level synchronization, EgOmini 4X reliably outputs frame-aligned, egocentric multi-view video, ideal for wearable capture, robot-mounted capture, lab scene playback, and other data collection workflows.

  • A near-ring multi-camera layout covers forward, side, and partial downward views, preserving more complete first-person task context.
  • Synchronized capture combined with local recording reduces the cost of downstream registration, clipping, and pre-training processing.
  • TF / SD local recording, USB offload, and Wi-Fi + BLE management form a complete workflow.

PRODUCT PARAMETERS

CameraRES @ FPSEncodingFOV (D)FOV (H)FOV (V)Distortion
Camera 11080p @ 30fpsH.264 / H.265194° ±5152.6° ±3121° ±3-100% ±5% (@180°)
Camera 21080p @ 30fpsH.264 / H.265187.6° ±5156.4° ±383.4° ±3-100% ±5% (@180°)
Camera 31080p @ 30fpsH.264 / H.265145° ±5120.6° ±364.2° ±3-64.6% ±5%
Camera 41080p @ 30fpsH.264 / H.265200° ±5200° ±5108.2° ±3-100% ±5% (@180°)

Note: The Y value in parentheses after FOV is the corresponding image-height reference; distortion values are calibrated at the indicated field of view. All parameters are subject to measured calibration; ± indicates the engineering tolerance range.

  • Ring-Shaped Wide-Angle Layout

    Four distinct wide-angle modules form near-ring coverage around the body, more fully preserving the operator's viewpoint, environmental context, and partial downward views in first-person tasks.

  • Built for Fast Motion

    Global shutters and microsecond-level synchronization reduce distortion and timing errors during rapid motion and view switching, making the device well suited to real-world motion capture and pre-training processing.

PRODUCT HIGHLIGHTS

  • 01QUAD CAMERA ARRAY

    Surround Multi-Camera Capture

    • Four wide-angle modules form a near-ring field of view covering forward, side, and partial downward areas.
    • Preserves more of the first-person action and its environmental context, minimizing missing key viewpoints.
  • 02µs-LEVEL SYNC

    Closed-Loop Microsecond Sync

    • Unified triggering, recording, and time synchronization keep all views frame-aligned.
    • Reduces data cleaning, time alignment, and post-processing costs, fitting right into training data production.
  • 03COMPACT DEPLOYMENT

    Compact Deployment & Management

    • Master-slave coordination and centralized power suit wearable use, robot mounting, and rapid lab deployment.
    • Wi-Fi + BLE management reduces on-site operational complexity.
  • 04LOCAL DATA PIPELINE

    Reliable Data Pipeline

    • TF / SD local recording, USB offload, and on-site verification form a complete capture pipeline.
    • Balances stable continuous capture with efficient data retrieval across experiment batches.

APPLICATION SCENARIOS

  • Imitation Learning Data Capture

    Imitation Learning Data Capture

    • Records human operations from multiple first-person views, fully preserving the relationship between hand movements and target objects.
    • Ideal for task demonstrations such as grasping, assembly, and tabletop interaction.
    • Synchronized multi-view footage reduces downstream registration and clip-selection costs.
  • Visual Navigation & Embodied Perception

    Visual Navigation & Embodied Perception

    • The near-ring multi-camera layout preserves more complete spatial context, path information, and occlusion changes.
    • Suited to continuous capture on mobile robots, wearables, and in complex environments.
    • Ideal for building embodied perception datasets for indoor navigation, inspection, and human-robot collaboration.
  • World Model & Multimodal Training Data

    World Model & Multimodal Training Data

    • Synchronized multi-camera capture and local storage provide stable input for temporal prediction.
    • Supports world model training and multimodal data organization, making it easy to accumulate continuous observation sequences.
    • Helps build a unified data foundation for perception, prediction, and decision-making tasks.