Skip to content
Spatial AI · Spatial Intelligence

Many cameras. One space.

We align anonymized signals into one coordinate system and journey.

01

Signal

Anonymized signals

02

Coordinate

Aligned to one coordinate system

03

Journey

One journey

MTMC merges observations from many cameras into one spatial coordinate system.

Journey Proof

A journey is an event record, not a single frame

Connect observations scattered across cameras on the same coordinates, and everything from entry to exit stays as one record.

One journey on shared coordinates

  • Customer
  • Staff
  • Trajectory

Event 1 / 4 · 10:02 Entrance crossed

Anonymous event log

Select an event to highlight that point on the map.

Connecting events in one coordinate system lets the team review movement, dwell, and exit as one flow.An example using illustrative observations and a same-seed simulation.

Zone classification derived from the same observation

The baselines are the medians across all zones, drawn on the chart. Each quadrant follows from the visit and dwell coordinates of the zone.

Illustrative simulation in a virtual space. Values do not represent actual performance.

Zone quadrant matrix

Visits and average dwell per visit sort every zone into four types. The baselines are the medians across all zones.

↑ Avg. dwell per visit (sec)Visits →Median visits 430Median dwell 2.5sHigh visits · high dwellHigh visits · low dwellLow visits · high dwellLow visits · low dwellABCDE
  • ALow visits · high dwellvisits 96 · dwell 10.1s
  • BHigh visits · low dwellvisits 573 · dwell 1.4s
  • CLow visits · high dwellvisits 96 · dwell 2.5s
  • DHigh visits · high dwellvisits 430 · dwell 5.1s
  • EHigh visits · low dwellvisits 478 · dwell 1.3s

AI contrast

Same question, different answer

An AI that cannot see the space stays with generalities. With coordinates and journeys, SAAI answers with the numbers of this store.

Same Question

"Yesterday evening, which product drew the longest dwell?"

Generic AI chatbot
Type a message…

* Both screens show simulated demo data for a virtual store.

Spatial capabilities

Read signals in the language of space

The four capabilities behind that difference.

01 · Computer Vision

Signals from one feed

  • People, pose, and density
  • Gaze and fixture contact
person-detectpose-estimatereid-embedmtmc-trackshelf-state

Anonymize · recognize · space · flow · change — a digest of the 5-category model catalog.

02 · MTMC

Cameras into one coordinate system

  • One flow across cameras
  • Journeys, dwell, and density on one map
PixelsCameraCoordinates
03 · VLM

A scene as a question

  • Describe what is happening
  • Read context beyond coordinates
04 · Domain Knowledge

Spatial signals for operations

  • Interpret signals by space
  • Connect them to the next decision

Why This Is Possible

Anonymization comes first

Identity disappears on input. Only the signals needed to read a space continue downstream.

More on the anonymization technology

Faces erased, context preserved

Identity is permanently erased, retaining only minimum features needed for scene understanding.

Faces masked as noise. Still readable — age, gaze, speech. Zero identity.

References

DEEPINGSOURCE, refining video anonymization technology since 2018, has presented and validated spatial intelligence at CES, in collaboration with KDDI, and within the NVIDIA ecosystem. Relevant presentation and collaboration records are available as reference materials.

Set the standard for reading your space

Review camera placement and analysis goals with our technical team.