Know what you’re evaluating.

Current workspace and source handling

The hosted workspace presents analyzed recorded replay with precomputed generic person and selected vehicle detections, temporary track IDs and source-linked observations. The selected sample determines which observations are available. It does not ingest live cameras.

Webcam or RTSP camera inputs and local deployment require a separately agreed scope and compatibility assessment.

Review decisions and notes are stored in the current browser. They are not shared, durable or tamper-evident audit logs. Clearing browser storage can remove these local records.

Capability and evidence boundaries

Implemented observations cover people and selected vehicle classes. Configured entry/exit, directional crossing, observed dwell, sustained occupancy/count and image-space person–vehicle proximity review cues require suitable fixed-camera footage and site-specific tuning. These rules have not been validated for a production deployment.

Two-dimensional proximity is an image cue, not a measured physical distance or collision prediction. A cue is not proof of an incident, rule violation or unsafe behaviour.

PPE, smoke/flame, fall/person-down, number-plate reading, queue analysis, violence, abandoned-object, industrial-fault and facial-recognition models are not active included capabilities. Specialist use cases need separately scoped models, suitable data, temporal evidence and measured evaluation. Queue analytics, shared audit records and production camera-fleet management are not demonstrated features.

Workspace controls and exports

The review workspace supports recorded-source selection, playback and seek controls, replay speed, track and zone overlays, an event timeline, and evidence-inbox filtering by event type and review status. Available source-linked MP4 excerpts can be downloaded with their attribution. A structured JSON review export includes the source, run, event and review context.

Rule setup adjusts a configured image region, duration and occupancy threshold against existing analyzed tracks. It does not rerun the detector or establish a live camera connection. Notes and decisions remain browser-local unless exported.

Sample coverage

The five Indian samples show roads and crowds. They do not validate industrial performance. Fixed-zone, dwell and crossing rules are disabled for those moving views. The supplemental factory clip has unverified geography. No industrial accuracy, response-time or savings claim is made.

Website sector imagery is AI-generated and illustrative. It is not customer footage, a product screenshot or evidence of a deployment. No endorsement or customer relationship is implied.

Security controls and data planning

The hosted application includes password-hash authentication, private media storage, authenticated asset routes, and CSRF/session controls. HTTPS is part of the hosted deployment. These concrete controls are a baseline, not an independent security certification or a complete enterprise control framework.

Define these requirements for your engagement

• Monitoring purpose, footage permission and the people responsible for review.
• Hosting environment, network access, data location and any processing outside the site.
• Retention, deletion, backups, incident response and secure transfer arrangements.
• User roles, access lifecycle, shared review history and audit requirements.
• Applicable privacy, employment, procurement and sector obligations.

On-premise/private-cloud choices, SSO, role-based access, shared audit trails, custom integrations and alert delivery are scoping options, not assumed included features. Company credentials in the Any Tech Ventures Pvt. Ltd. corporate footer do not certify this product. No blanket legal-compliance claim is made.

The website’s enquiry choices prepare an email in your email app. They are not submitted to a backend, stored by the page or sent to tracking services. Share only a high-level requirement initially. Agree a secure method before transferring footage.

A useful evaluation starts with an agreed question

1. Choose a priority: define the operational question and the action a human reviewer would take.
2. Assess representative footage: include real angles, lighting, occlusion and normal activity from a source you are authorised to use.
3. Configure deliberately: use suitable fixed views for zones, crossing, dwell and count cues.
4. Measure together: review missed cues and false cues against agreed acceptance criteria. Do not substitute a polished demonstration for site validation.
5. Agree rollout boundaries: document data handling, review ownership, support, integration and the commercial scope.

The demonstration is not a safety-certified system. A missed event does not establish that a scene is safe. Keep original footage available and accountable human judgment in the workflow. Ai Drishyam does not provide personal profiles, intent/emotion scoring, automated penalties, disciplinary decisions or equipment controls.