Krishna WiFi
Multi-tenant web panel to configure and manage MikroTik routers remotely over RouterOS API — no WinBox needed.
AI-powered forensic CCTV analysis for investigators
When an incident is reported, investigators inherit a brutal task: dozens of hours of CCTV footage from multiple cameras, and somewhere inside it, a few seconds that matter. Watching it all manually takes days of an officer's time and still misses things at 2 AM on the second screen. Krishna Intelligence is forensic CCTV video analysis software built for exactly this work. The Windows desktop application scans footage in batches, using YOLOv8 detection and tracking — with automatic GPU acceleration — to find and follow every person, vehicle and animal across the videos. It captures one enhanced evidence photo per unique object, so instead of 14 hours of video, the investigator reviews a clean gallery of every distinct subject that appeared, each timestamped and traceable to its source frame.
The analysis toolkit reads like an investigator's wishlist. Smart filters narrow results by object type and colour; clothing colour is split into upper and lower body, so "blue shirt, black pants" is an actual query. The software estimates direction of movement and approximate relative speed, raises loitering and crowd alerts with adjustable thresholds, and suggests cross-video matches so the same suspect can be followed across cameras within a case. Face recognition runs against a local known-suspects folder, number-plate OCR reads vehicle registrations, and overlay-timestamp OCR recovers the camera's own burned-in clock. Live RTSP/HTTP scanning extends the same engine to real-time monitoring. Every saved evidence image can be face-blurred and watermarked, and all evidence is encrypted on disk.
Where Krishna Intelligence differs most from generic video tools is its chain-of-custody discipline. Case reports are never saved on the analysis PC — only evidence photos are — and are published exclusively to the secure web panel, where they open only via secret token links; the operator receives the view link on WhatsApp, and offline reports queue and upload automatically when connectivity returns. PDF exports carry a QR code linking to the online report, a SHA-256 verification code, the operator's name and designation stamp, the police-station header pulled from the operator's account, and a SHA-256 hash of every source video, making tampering detectable. The web panel adds admin-controlled user validity, WhatsApp OTP self-registration with admin approval, full audit logs, login history down to PC name and IP, station-wise statistics, CSV export, bulk ZIP download — and a Gujarati help guide for local teams.
Queue entire folders of CCTV footage and let YOLOv8 detect and track persons, vehicles and animals, using the GPU automatically when available. Days of manual footage review compress into hours of automated, repeatable analysis.
Filter every detection by object type and colour to isolate exactly what the case needs — a white car, a red-shirted man. Investigators query footage like a database instead of scrubbing timelines frame by frame.
The engine captures a single, best-quality photo of each unique person, vehicle or animal, with enhancement and night-vision processing applied. Reviewers face a clean gallery of distinct subjects instead of thousands of near-duplicate frames.
Clothing colour is analysed separately for upper and lower body, matching how witnesses actually describe suspects. A description like "blue shirt, black trousers" becomes a direct, executable search across the entire case.
Each tracked object carries its direction of movement and approximate relative speed. Investigators can reconstruct routes, test witness statements against physics, and separate a casual passer-by from someone fleeing the scene.
Adjustable thresholds flag subjects lingering unusually long or crowds forming in the frame. Pre-incident behaviour — the recce before the theft, the gathering before the clash — surfaces automatically instead of being noticed only in hindsight.
Within a case, the system suggests when the same person or vehicle appears across different videos and cameras. A suspect's full movement across a market or campus assembles itself from fragments no human reviewer would connect.
Faces detected in footage are compared against a local known-suspects folder that never leaves the machine. Repeat offenders are flagged in minutes, while the sensitive reference data stays entirely under the station's own control.
Saved evidence photos can automatically blur uninvolved faces and carry watermarks. Bystander privacy is protected and every image is marked traceably — essential when material may be shared with courts, lawyers or the press.
OCR reads vehicle number plates and also recovers each camera's burned-in overlay timestamp. Vehicles become searchable text, and events can be placed on the camera's own clock even when the DVR's metadata is wrong.
All evidence stored by the desktop application is encrypted on disk. If an analysis laptop is lost, stolen or seized, case material remains unreadable — protecting both the investigation and every person appearing in the footage.
Connect live camera streams over RTSP or HTTP and run the same detection engine in real time. The tool that reconstructs yesterday's incident can also watch today's sensitive location as events unfold.
PDF reports include a QR code linking to the online report, a SHA-256 verification code, the operator's name and designation stamp, the police-station header from the operator's account, and a SHA-256 hash of every source video. Any tampering with report or footage becomes provable.
Admins create users with office, designation and WhatsApp number, set per-user validity in days, and get report search, WhatsApp resend, CSV export, bulk ZIP download, report sharing, station-wise and monthly statistics, a full audit log, login history (PC name, OS, PC user, IP), WhatsApp-OTP self-registration with admin approval — plus a Gujarati help guide.
Run the standalone Windows EXE — no complicated setup — and log in with the account your admin created with your office and designation.
Start a case and queue CCTV files in batch; YOLOv8 scans them with automatic GPU acceleration while you continue other work.
Review one evidence photo per unique object, filter by type, colour and clothing, follow cross-video matches, and run face and plate recognition.
The report publishes online only — never saved on the PC — and the view link arrives on your WhatsApp, with a verifiable PDF export on demand.
Turn seized DVR footage into court-ready, hash-verified evidence reports in hours, with station headers and operator stamps built in.
Run multi-camera, multi-day analyses with cross-video matching and a full audit trail of who accessed which report and when.
Deliver professional incident-reconstruction reports to clients, backed by verifiable hashes rather than screenshots and opinion.
Search long surveillance recordings by clothing colour, vehicle or plate number instead of spending billable days watching footage.
Investigate thefts and incidents across large campuses fast, and use loitering and crowd alerts on live cameras for prevention.
Pay once, use all year. No per-order cuts, no hidden charges.
By design, at every layer. Reports are never saved on the analysis PC — only evidence photos are, and those are encrypted on disk. Reports publish exclusively to the web panel and open only via secret token links, with the view link delivered to the operator on WhatsApp. Every PDF carries a SHA-256 verification code, a SHA-256 hash of each source video, the operator's name and designation stamp and the station header — so any later tampering with the report or the footage is provable, not just suspected.
Analysis runs fully offline — YOLOv8 detection, tracking, filtering, face recognition against the local suspects folder, OCR and evidence capture all happen on the local machine, which matters for sensitive material. Internet is needed only to publish the case report to the web panel. If connectivity is unavailable when analysis finishes, reports queue locally and upload automatically the moment the connection returns, with the WhatsApp view link sent at that point. Field work in poorly connected areas is a normal, supported scenario.
It is built for real Indian CCTV, not laboratory conditions. Common DVR and NVR export formats are supported for batch analysis, and live RTSP/HTTP streams can be scanned directly. Enhancement and night-vision processing improve dark or grainy frames before evidence capture, and overlay-timestamp OCR recovers the camera's burned-in clock even when DVR metadata is missing or wrong. Detection quality naturally rises with resolution, but the pipeline is tuned to extract usable evidence from typical low-cost camera footage.
Face recognition compares faces detected in footage against a known-suspects folder stored locally on the analysis machine — images the investigating unit itself maintains. Nothing is matched against any cloud database, and the reference folder never leaves the PC, keeping the unit in complete control of this sensitive data. Matches are presented as investigative leads for human confirmation, not automated verdicts. For bystanders, automatic face blurring on saved evidence protects the privacy of people unconnected to the case.
Through the web panel, admins create each operator with their office, designation and WhatsApp number, and set an account validity in days — access expires automatically when a posting ends. Operators can also self-register with WhatsApp OTP, pending admin approval. Every action lands in a full audit log, and login history records PC name, OS, PC username and IP address. Station-wise and monthly statistics, report search, CSV export and bulk ZIP download give supervisors complete oversight, with a Gujarati help guide for local teams.
Multi-tenant web panel to configure and manage MikroTik routers remotely over RouterOS API — no WinBox needed.
Visitors register on WhatsApp, receive a designed QR pass card, and are scanned in at the gate — one platform, end to end.
All-in-one WhatsApp automation for Indian businesses — bulk messaging, AI chatbot, IVR menus and API in one dashboard.
Real situations with full setup guides — find yours.
Cloud software — works in every city, with WhatsApp setup help in Gujarati and English.
AI-powered forensic CCTV analysis for investigators — see it live right now, or message us and we will set it up for you.