INDUSTRIAL EMBODIED AI DATA SYSTEM
Capture. Synchronize. Process.
Prepare data for embodied AI.
01 — SYSTEM ARCHITECTURE
FMC³ Doceo System connects multimodal data capture, synchronization and processing in one integrated workflow, turning real-world human demonstrations into structured data for embodied AI development.
Open full-size architecture 02 — DATA PROCESSING WORKFLOW
The workflow preview brings aligned RGB, depth, segmentation, hand-pose and sensor streams together in one processing view.
03 — MEET FMC³ AT DMTA
DEEP DIVE · TECHNOLOGY & ENGINEERING
A practical walk through the data pipeline—from acquisition and processing to model training and transfer to a real robot.
Dr. Joachim Fetzer · Prof. Björn Giesler
LIVE DEMONSTRATION · MOBILE ROBOTICS
See how human-centric data capture supports a live task demonstration and robot reproduction on FMC³ Franzi.
Hands-on demonstration
USE CASE WORKSHOP · INTEGRATION
A grounded discussion of where humanoids fit, where mobile Physical AI systems may be better, and when the process itself should change first.
Dr. Joachim Fetzer · Florian Obermeier · Prof. Björn Giesler
04 — DOCEO REALCAP HARDWARE
Doceo RealCap combines first-person vision, hand pose, tactile force, motion and synchronized local recording. The modules can be configured around the task and the required data objective.
REALCAP 01 — HEAD-MOUNTED CAMERA
A six-camera global-shutter vision system integrating edge computing, wireless connectivity, storage, IMU, audio and battery power for synchronized mobile capture.
REALCAP 02 — PORTABLE DATA CAPTURE BOX
Two portable configurations centrally manage cameras, data gloves, live preview, calibration, local storage and data upload. Select RK3576 or RK3588 according to compute, interface and recording requirements.
REALCAP 03 — DATA GLOVES
Hand data combines pose and contact mechanics. Bending-sensing and tactile gloves can be used independently or together according to the training objective.
Bluetooth wireless and USB connectivity.
Hand-pose sensing range across finger joints.
Hand-pose accuracy for finger joints and finger-web stretches.
Hand-pose sensor durability with resistance change below 8%.
Configure the glove for each data objective.
In addition to pressure detection, provides decoupled bending sensing; a single tactile glove can output pressure and pose simultaneously.
Integrated into the glove, providing complete global hand-motion data.
Customizable battery life enables a fully wireless configuration.
Upgrades tactile coverage from 400 to 600 sensing points.
Adds pressure points to the sides of the fingers at the same density as the front.
Adds one IMU to each fingertip for high-accuracy fingertip orientation and per-finger motion tracking.
REALCAP 04 — WRIST-MOUNTED CAMERA
The wrist camera complements the head-mounted global view with a stable perspective of the operating area—even when the hand is partially occluded by the body or workpiece.
05 — DOCEO REALCAP DATA OUTPUT
Streams are organized by schema version in meta.json, with Python and C++ parsing examples for integration into robot-operation training workflows.
1600 × 1200 · 30 fps
RAW / YUV1920 × 1080 · 30 fps · 130°
MJPEG → MKVAccelerometer + Gyroscope
CSV19 points + optional 9-axis IMU
CSV400 / 600 points · 200 / 400 Hz
Binary (.bin)Nanosecond accuracy
JSONLIntrinsics · Extrinsics · IMU bias
JSONProduct specifications are subject to change without prior notice. Please refer to our latest documentation.
We assess how our solution fits into and improves your existing operations.
We test the solution in your workflow and demonstrate measurable results.
We work with you to scale performance and support long-term operational success.