3TIUM Installed

Software is installed on a computer or a device,
taking full advantage of available hardware.

Browser Only 3TIUM

Software runs purely on a web page without installation,
with container limiting performance and quality.

• Platform: Windows, WinCE, Linux, Android, FreeBSD, OS X, iOS
• Programming Language: C++, Java, openCL, assembler
• Graphics: openGL, DirectX; Maya and 3ds Max model conversion

• Platform: Flash or HTML5
• Programming Language: Action Script(AS3), javascript
• Graphics: Hardware Accelerated Flash, Maya model import

• Camera Type: The better, the better. Better optics will contribute to precise position and orientation estimation. Higher resolution, better sensitivity and less sensor noise provides bigger operable range and tolerance towards operating environment. Better frame rate will capture quicker moves. For massive use interactive applications almost any camera will do. For high precision applications we make sure you have the right camera with our system.
• Supported Modes: 160X120 30fps; 176X144 25fps; 320X240 30/60fps; 352X288 25fps; 640X480 15/30/60fps; 720X480 30fps; 1280X720 15/30/60fps; 1600X1200 30fps; 1920X1080 15/30/60fps.
• Frame Encoding: YUV, RGB, JPEG
• Recommended Minimal Field Of View: 44° top-to-bottom and 56° left-to-right

• Recommended Modes: 352X288 25fps, 640X480 15/30fps, 1280X720 15fps.
• Frame Encoding: YUV, RGB, JPEG
• Recommended Minimal Field Of View: 44° top to bottom and 56° left-to-right
• Recommended application: auto-pilots, medical vision and factory robots.
• Features:
  • Simple 3D space reconstruction by triangulating calibrated stereo input.
  • Efficient 3D space reconstruction of still content.
  • 3D position and rotation estimation of low feature density objects.

• Infrared laser beam may be detrimental to your eyes. We do more on a simple camera than the depth camera will ever do. But if you insist, we are experienced with the SDK from Habarzel street.

• Operation Range:
  • 0.5m-10m at 1920X1080px
  • 4cm - 0.8m at 160X120px.
• Adaptive Tracking Precision:
  • Horisontal: 0.01-3.3px/mm
  • Vertical: 0.015-2.5px/mm
  • Depth: 0.05-2px/cm
• Average Deviation:
  • Horizontal/Vertical: 0.4%
  • Depth: 2.5%
• Tracking speed (@30fps):
  • 24m/s
• Hardware: 1GHz Dual Core CPU, 512MB RAM

• Operation Range:
  • 0.5m-10m at 1920X1080px
  • 4cm - 0.8m at 160X120px.
• Adaptive Tracking Precision:
  • Horisontal: 0.01-3.3px/mm
  • Vertical: 0.015-2.5px/mm
  • Depth: 0.05-2px/cm
• Average Deviation:
  • Horizontal/Vertical: 10%
  • Depth: 33%
• Tracking speed (@30fps):
  • 2.7m/s
• Hardware: 1GHz Dual Core CPU, 512MB RAM

• Object Recognition Time: 100ms*
• Recognition Hit Rate: 95%
• False Acceptance Rate: 1%
• Average object tracking cycle: 2ms*
• Maximum Camera Tilt (for flat marker-free images and surfaces): 60°
• 3D Position Estimation Slip**: 0.7% of augmented range.
• 3D Estimated Object Orientation Accuracy(XYZ)**: ±3° to ±5°

* On 1GHz dual-core CPU, 512MB RAM and 640X480 capture mode.
** For objects bigger than 5% of the camera view in low optical distortion area.

• We are not offering Shambala for in-browser applications at the moment. But give us a good reason and we will.

• Performance:
  • Video/Image Indexing Speed*: 8MB/s
  • Unindexed Content Search*: 11MB/s
  • Indexed Content Search*: 5s
  • Average multimedia OLAP Report*: 23s
  • Index Footprint: ~0.8% of raw data
• Recognition Quality:
  • Recognition Hit Rate: 95% at 1% FAR; 97% at 5% FAR
• Cloud Processing: Enabled
• Cloud Data Clustreing: Horizontal and Hierarchical.
• Data Mining: Analytics, Ratings, Trends, Optimization, Behavior Modeling.

* On Server with 2.2GHz Quad Core CPU, 4GB RAM, RAID storage, test set: 1TB raw multimedia content.