Xnap Box firmware 0.9.8 user manual is out

2 major features : Tracker ID and Sharpness sampling are now available in Xnap Box firmware 0.9.8. Please check the Quick Start Guide from User Manual page of this blog.


Firmware version 0.9.8 Xnap Box is out!

A major upgrade of Xnap Box firmware is just out!

Xnap Box datasheet en 20170726

Salient features of Xnap Box firmware version 0.9.8.

  • Sample selection
    • blur filtering threshold
      – filter out face/object image with sharpness index which is below your desired level (in red).
    • best sharpness sampling
      – choose the top (n) number of images in good sharpness index among the “sampling window size”.  This option is mutually exclusive with “best dimension sampling”.
    • best dimension sampling
      – choose the top (n) number of images in largest size among the “sampling window size”.  This option is mutually exclusive with “best sharpness sampling”.
  • Introduction of Tracker ID
    • each person appearance will be tracked and assigned with Tracker ID: a unique ID sticky throughout the life cycle of the given face.
    • No. of image images to be exported can be configurable based on sharpness/face size/sample parameters
  • Introduction of Live View
    • Export of face-annotated “Live View” to assist configuration or for display on application (overlay sharpness & Tracker ID)
  • Detection Strength configuration
    • to enable the administrator to configure the robustness of face finder
  • Improved System Status page
    • to provide more readable system operating status messages of Xnap Box assist troubleshooting and performance finetune

The new firmware requires 5V 3Amp micro-USB power supply to power up the Xnap Box with horsepower to run new extra features of Xnap Box properly. Please check the power specification of the power supply on your hand before contemplating firmware upgrade of Xnap Box.

Xnap Box configuration page

Xnap Box Data Grabber SDK

Developers would ask: how can I integrate Xnap Box with my existing application? How can I grab the cropped photos of faces/objects being detected by Xnap Box? Besides of image file, how can I get the meta of these images, such as timestamp, the dominant color of objects, Tracker ID (upcoming) and so forth?

Xnap Box Data Grabber SDK is now available on GitHub.


Currently, Xnap Box Data Grabber SDK comes with 2 different favors:

Java SDK




Python SDK (upcoming)



Xnap Box Data-sheet (0.9.7)

Xnap Box datasheet can be downloaded here:

Xnap Box datasheet en 20170308

Key Features:

  • Each Xnap™ Box processes one RTSP H.264 / MJPEG video stream over IP network, or MJPEG AVI USB File sources on FAT32 thumb drive
  • Support up to 1920×1080 resolution with precise resolution auto-detection
  • Near real-time face detection algorithm using deep learning technology
  • Export MJPEG face/object streams, alongside with meta-data, through HTTP protocol by default
  • Optional: face/object image output to USB thumb drive
  • Detect multiple faces and objects with size range from 72 to 1200 pixels in 1080p (2 mega pixel) frame
  • 20º yaw/pitch/roll pose variation
  • Face detection missing rate* < 0.01%
  • Compact-size enclosure
  • Powered by 5V 2.5A Certified MicroUSB cable
  • Bonjour Support for the latest web browser
  • Built-in object color detection to provide dominant color image of pedestrian/vehicle (meta-data to be sent via HTTP)
  • NTP support for time synchronization to ensure precise (up to microsecond level) timestamp with all captured images
  • Filename of face image shall be derived from camera system time (via ONVIF) & timestamp information of the concerned frame of the input RTSP source (only in local storage mode)
  • Simple web-based guided operation to export face images to USB thumb drive
  • Built-in web-based utilities to fix/format USB thumb drive into FAT32 / EXT4 format
  • 802.11n Soft-AP mode support to allow web-based administration from any PC / smartphone/tablet with a standard web browser.

As a prelude, we are planning to add these 2 features in next release

  • Administration web with password authentication
  • Built-in ONVIF discovery tool
  • Built-in face blur filter to reduce undesirable blurry face images

HTTP Headers from XB Object/Face Streams Spec

From version >= r.0.9.8


YYYYMMDD=Year,Month,Day (system wide)

HHMMSS=Hour,Minutes,Seconds (system wide)

SSSSSSSSSSS=stream time in seconds (per session)

FFFFFFFF=frame no/count (per session)


(Object/Face Centroid X in the whole frame, integer: 0-1200)

X-objectXpos: 9999

(Object/Face Centroid X in the whole frame, integer: 0-2000)


X-objectWidth: 9999

(Face Width in XB Face, integer: 72-1200)

X-objectHeight: 9999 (Face Height in XB Face)

(Image width & height in Object, Face Width & Height in XB Face, integer: 72-1200)


X-TrackerID: 99999

(Integer from 0-65535, back to 0 after 65535)

X-TrackDir: (obsoleted)

X-ObjectColor1HSV: #999#999#999

(Dominant Color, H, integer: 0-360, S, integer: 0-100%, V, integer: 0-100%)

X-ObjectColor2HSV: #999#999#999

(2nd Dominant Color, H, integer: 0-360, S, integer: 0-100%, V, integer: 0-100%)

Already exists since version 0.4:

Server: (XnapBOX machine name)