See
Every frame understood
The model describes what passes: vehicle type, colour, plate, and the text and logos on the body.
Tell us what you're evaluating and our team will get in touch.
We've received your request. Our team will reach out shortly.
Share your details and we will start the download.
Thanks. Your download should begin shortly. We may follow up about the programme.
Download againShare your details and we will start the download.
Thanks. Your document should begin downloading. Our team may follow up if useful.
Download againAsk in natural language. Get the evidence.
AI Search runs a vision language model on the camera itself. Describe the vehicle or the person you are looking for the way you would say it out loud, and get matching frames with evidence instead of hours of footage to scrub.
Conventional review means choosing a camera, picking a time window, and scrubbing through hours of footage. AI Search flips that: the camera already understands what it saw, so you ask in natural language and get matching evidence back.
The model describes what passes: vehicle type, colour, plate, and the text and logos on the body.
Your description is matched against those understandings, so a phrase works where a dropdown cannot.
Each match returns the frame, the timestamp and the plate.
Forensic review usually means choosing a camera, narrowing a time window, and scrubbing through hours of footage. AI Search inverts that: describe what you saw and get matching evidence back.
Pick a camera, set a time range, scrub frame by frame — and hope the moment is still in the recording.
Describe what you saw in your own words. The camera returns ranked matches with frame, timestamp and plate.
Pick a camera, set a time range, scrub frame by frame
Type a plain-language description of what you are looking for
Plate number, lane and approximate time before you can start
A visual description — colour, sticker, clothing, damage, vehicle type
Hours of manual review across multiple cameras
Review ranked matches in seconds, not shifts
Metadata filters only — plate, time, camera
Type, colour, text, logos, stickers and people attributes
Often server-side; footage may need uploading first
On-device VLM on Recon — query answered at the gantry
Manual clip export if you find the right moment
Frame, timestamp and plate returned with every match
Repeat the same process per camera and per console
EdgeTrack extends one query across every connected unit
Investigations move faster when operators describe what they saw instead of replaying everything the camera recorded. These are the outcomes teams see once plain-language search is live.
Turn hours of scrubbing into seconds of review. Operators start with matches, not raw footage.
Describe what you saw — a sticker, a jacket, a vehicle colour — and let the camera find candidates.
Go beyond dropdown filters. Search logos, text on glass, clothing, damage and vehicle attributes.
Each result returns the frame, timestamp and plate — ready for enforcement or dispute resolution.
The model runs on the camera. Queries are answered locally and footage never has to leave the site.
EdgeTrack extends the same query across every connected unit — one description, many cameras.
These are descriptions, not filters. Nobody has to know the plate, the lane or the minute to start looking.
Click any example. The matching evidence image appears on the right.
AI Search comes with Recon Global Shutter.
The model sits on the unit, so a query is answered on site and footage never has to be uploaded.
EdgeTrack extends the same query to the connected units, turning one camera into a corridor.
AI Search is an advanced search feature that uses a Vision Language Model (VLM) to let operators query camera feeds in natural language. Instead of scrubbing hours of footage, you type a description and get matching results in seconds.
Text on vehicles (e.g., "Ganesh" written in Marathi), brand logos (e.g., "Puma"), appearance attributes (e.g., "man with red cap," "yellow face emoji sticker"), vehicle descriptions (e.g., "damaged white truck"), and partial plate matches.
English, Hindi, and Marathi. The system recognizes text in all three languages on vehicles, signs, and other surfaces.
Results are returned in real time with minimal latency. The search results update dynamically as new data is added from the camera feed.
AI Search runs on the Recon Global Shutter camera's edge processor (157 TOPS, 1024 CUDA cores). No server-side processing is needed for the search itself, though results are viewable on the EdgeTrack dashboard.
Law enforcement investigations (finding suspect vehicles or people by description), highway monitoring (finding specific vehicles across corridors), and forensic analysis (searching recorded footage for specific events or objects).
AI Search is an optional add-on to the Recon Global Shutter camera. It can be enabled on any deployed camera without hardware changes.
Traditional ANPR search is limited to plate numbers. AI Search goes beyond the plate. You can search by vehicle appearance, text, logos, human attributes, or any visual detail the camera has captured. It answers "what did you see?" not just "which plate passed."