# AiVisionBot

> AiVisionBot is the AI computer vision service of VYROX AI (Kuala Lumpur, Malaysia), powered by AiServa. It turns ordinary CCTV and IP cameras into systems that detect, recognize, analyze, process and inform. The site shows 39 use cases, each with a 3D demo (a simulation for explanation, not model output) and 58 generated illustration photos in total with example detection boxes.

Last updated: 2026-10-19

## Key Facts
- Company: VYROX AI, Kuala Lumpur, Malaysia. Serves Malaysia and Southeast Asia.
- Contact: WhatsApp Patrick Au +60 19-688 3338 (https://wa.me/60196883338). No email contact is published.
- Team: Ts. Dr Leong Yee Rock (AI Specialist and Forward Deployed Engineer); Au Sek Howe (AI Specialist and Forward Deployed Engineer); Teh Hon Kee (AI Specialist and Forward Deployed Engineer).
- How it works, five steps: Detect: Every frame is scanned for the objects the use case needs. Recognize: Objects get a class, a track ID, a pose, a reading or a code. Analyze: Counts, zones, dwell time, speed, payload and trends are worked out. Process: Events are confirmed over time so a single odd frame never raises an alarm. Inform: People and systems are told: WhatsApp, dashboards, webhooks, MQTT, relays.
- Software: the VYROX Vision Engine, ten proprietary engines: VYROX HawkEye (Object Detection); VYROX OmniSight (Open Vocabulary); VYROX PixelBlade (Segmentation); VYROX TrailLock (Tracking); VYROX VisionMind (Vision Language); VYROX MotionIQ (Pose and Action); VYROX ReadSight (OCR); VYROX FlawHunter (Anomaly); VYROX EdgeCore (Inference); VYROX ModelForge (Data and Training).
- Deployment: inference can run on an edge box at the customer site, sending only events, snapshots and counts. Cloud is optional.
- Photos are generated illustrations, not customer footage. Detection boxes on photos are examples. 3D demos are simulations.

## Case Studies
- Quarry Crusher: Every Truck, Every Load (PROTOTYPE on a 3D digital twin; stage: Design done, Digital Twin done, Site Trial next, then Live). Problem: Nobody can say for sure how full each truck was, or whether every load really went into the crusher. Approach: Four CCTV cameras at the tipping point. Our own VYROX HawkEye model follows each truck from arrival to departure across 7 detection classes. Output: Payload as a percentage of rated volume with a range (a camera estimate, not a weighbridge), dump verification and a Person on Tipping Pad alarm.
- Rubber Glove Line Inspection (Manufacturing, reference design, not a customer result). Thousands of gloves a minute pass on ceramic formers, and manual spot checks miss pinholes and tears. Camera: Ring lit camera above each former line. AI: Flags pinholes, tears and stains. Output: Reject signal and a shift report with photos. Related demo: https://aivisionbot.com/#uc-qc-sorting
- Palm Oil Mill Bunch Grading (Agriculture, reference design, not a customer result). Grading fresh fruit bunches by eye at the ramp is slow and often disputed. Camera: Pole camera over the loading ramp. AI: Grades ripe, unripe, overripe and empty bunches. Output: Grade per lorry printed with the weighbridge ticket. Related demo: https://aivisionbot.com/#uc-tomato
- Broiler House Flock Monitoring (Livestock, reference design, not a customer result). Inactive birds and crowded or empty feeder lines are found on the morning walk, hours too late. Camera: Ceiling cameras along each house. AI: Counts birds and spots still birds and crowding at feeders. Output: Alert to the farm manager with the house and a photo. Related demo: https://aivisionbot.com/#uc-geese
- Sea Cage Feeding Watch (Aquaculture, reference design, not a customer result). Feeding is judged from the walkway, so feed is wasted or fish go hungry. Camera: Underwater and walkway cameras per cage. AI: Tracks fish activity and the feeding response. Output: Stop feeding signal and a daily activity report. Related demo: https://aivisionbot.com/#uc-koi

## Honest Limits
- Safety Aids, Not Guarantees: Fire, fall, drowning and weapon alerts assist people. They never replace certified alarms, lifeguards, carers or security staff, and false alarms or misses can occur.
- Measurement Needs Calibration: Level, volume and size claims need calibrated cameras and an uncertainty study.
- Camera, Lens and Light Matter: Small objects and fine defects need the right optics. We check on site before promising a number.
- Privacy by Design: Designed to support PDPA obligations with anonymous tracking, optional blurring and on site video. Compliance depends on each deployment and is reviewed with you.

## Pages
- [Home](https://aivisionbot.com/): use cases and 3D demos, how it works, capabilities, VYROX Vision Engine, case studies, honest limits, team, FAQ
- [Full detail](https://aivisionbot.com/llms-full.txt): every use case with its five steps, photos and engines

## Capabilities (12)
- Object Detection: Finds and labels every object in view.
- Object Tracking: Keeps one ID per object across frames.
- Counting: Counts items, animals or vehicles across a line or zone.
- Classification: Grades ripeness, quality, colour or condition.
- Segmentation: Marks exact pixels of cracks, smoke or water.
- OCR and Codes: Reads containers, meters, barcodes and labels.
- Pose Estimation: Finds body keypoints for falls, gestures and exercise.
- Action Recognition: Understands falling, tipping, smoking and drowsiness.
- Anomaly Detection: Learns normal and flags anything else.
- Thermal and Depth: Sees heat in the dark and measures distance.
- Visual Measurement: Estimates level, volume and size from calibrated views.
- Sound and Vision: Combines audio cues like a squeal or a cry with video.

## VYROX Vision Engine
- VYROX HawkEye (Object Detection): Nano, Swift, Pro. Best for Fast, accurate boxes on every frame.
- VYROX OmniSight (Open Vocabulary): Find anything you describe. Best for New objects described in words.
- VYROX PixelBlade (Segmentation): Pixel-exact masks. Best for Cracks, smoke, water lines, labelling.
- VYROX TrailLock (Tracking): One ID, start to finish. Best for Counting, dwell time, IDs through occlusion.
- VYROX VisionMind (Vision Language): Ask your cameras. Best for Grounded queries and incident descriptions.
- VYROX MotionIQ (Pose and Action): Body, face and hands. Best for Falls, gestures, drowsiness, exercise.
- VYROX ReadSight (OCR): Reads what it sees. Best for Containers, meters, labels, spines.
- VYROX FlawHunter (Anomaly): Finds the odd one out. Best for Defects with few bad samples.
- VYROX EdgeCore (Inference): Runs on site. Best for Many cameras per edge box.
- VYROX ModelForge (Data and Training): Learns from your site. Best for Labelling, error review, retraining on your site.

## Use Cases (39)

### 1. Dump Truck Payload and Dumping (Quarry and Heavy Plant, Prototype)
- Detects: Truck, Truck Tipping, Rock Flow, Body Load, Person, Carryback
- Detect: dump-truck
- Recognize: Knows the truck state: arrived, tipping, lowered, left.
- Analyze: Estimates payload as % of rated volume with a range and confidence (a camera estimate, not a weighbridge).
- Process: Locks the payload when the body rises and checks rock really flows into the crusher.
- Inform: Raises Person on Tipping Pad and Left Without Dumping, logs every visit.
- Engines: VYROX HawkEye, payload model, VYROX TrailLock
- Value: Every load accounted for at the crusher.
- Photo 1: https://aivisionbot.com/img/dt-tipping.jpg (example boxes: Hopper, Raised Body, Rock Flow, Truck Tipping)
- 3D demo: https://aivisionbot.com/#uc-dump-truck

### 2. Conveyor Counting (Manufacturing and Logistics)
- Detects: Biscuit, Count Line
- Detect: conveyor-count
- Recognize: Gives each item a track ID so it is counted once.
- Analyze: Counts items crossing the line and works out the rate per minute.
- Process: Compares the rate with the target and spots gaps or jams.
- Inform: Shows a live counter and alerts when the line slows.
- Engines: VYROX HawkEye Nano, VYROX TrailLock
- Value: Live output counts without a sensor per lane.
- Photo 1: https://aivisionbot.com/img/uc/conveyor-count.jpg (example boxes: Operator, Vision Camera)
- Photo 2: https://aivisionbot.com/img/uc/conveyor-count-2.jpg (example boxes: Biscuit)
- Photo 3: https://aivisionbot.com/img/uc/conveyor-count-3.jpg (example boxes: Biscuit)
- Photo 4: https://aivisionbot.com/img/uc/conveyor-count-4.jpg (example boxes: Biscuit)
- 3D demo: https://aivisionbot.com/#uc-conveyor-count

### 3. Quality Control Line Monitor (Manufacturing and Logistics)
- Detects: Gold, Silver, Red, Defect
- Detect: qc-sorting
- Recognize: Classifies the part by colour and spots surface defects.
- Analyze: Tallies grades and defect rate per shift.
- Process: Triggers the reject arm for a defective part.
- Inform: Logs each reject with a photo for the QC report.
- Engines: VYROX HawkEye Pro, classifier
- Value: Fewer escapes, every reject with evidence.
- Photo 1: https://aivisionbot.com/img/uc/qc-sorting.jpg (example boxes: Blue Disc, Operator, QC Monitor, Red Disc, Reject Bin, Ring Light Camera, Silver Disc)
- 3D demo: https://aivisionbot.com/#uc-qc-sorting

### 4. Particle Counting and Classification (Manufacturing and Logistics)
- Detects: Good Bean, Broken, Discoloured
- Detect: particle
- Recognize: Classifies each one as good, broken or discoloured.
- Analyze: Counts totals and the share of each grade.
- Process: Grades the batch against the buyer specification.
- Inform: Prints a grading report in seconds.
- Engines: VYROX HawkEye, VYROX PixelBlade
- Value: Fair, repeatable grading for every lot.
- Photo 1: https://aivisionbot.com/img/uc/particle.jpg (example boxes: Damaged, Discoloured)
- 3D demo: https://aivisionbot.com/#uc-particle

### 5. Barcode Detection and Reading (Manufacturing and Logistics)
- Detects: Barcode, Parcel
- Detect: barcode
- Recognize: Decodes the code and links it to the parcel.
- Analyze: Checks for duplicates and unreadable labels.
- Process: Matches the code to the order.
- Inform: Flags missing or wrong items before dispatch.
- Engines: VYROX HawkEye Nano, decoder
- Value: Scan without stopping the line.
- Photo 1: https://aivisionbot.com/img/uc/barcode.jpg (example boxes: Barcode Read, Operator, Parcel, Scanner)
- Photo 2: https://aivisionbot.com/img/uc/barcode-2.jpg (example boxes: Barcode, Parcel)
- Photo 3: https://aivisionbot.com/img/uc/barcode-3.jpg (example boxes: Barcode, Parcel)
- Photo 4: https://aivisionbot.com/img/uc/barcode-4.jpg (example boxes: Barcode, Parcel)
- 3D demo: https://aivisionbot.com/#uc-barcode

### 6. Container Number Recognition (Manufacturing and Logistics)
- Detects: Container, Code
- Detect: container-ocr
- Recognize: Reads the owner code, serial and check digit.
- Analyze: Validates the check digit and the ISO size code.
- Process: Matches the container to the booking at the gate.
- Inform: Opens the gate or holds it for a person to check.
- Engines: VYROX HawkEye Pro, VYROX ReadSight
- Value: Faster gate turnaround, fewer typing errors.
- Photo 1: https://aivisionbot.com/img/uc/container-ocr.jpg (example boxes: Container Code, Container Truck, Gate Camera, Gate Officer)
- Photo 2: https://aivisionbot.com/img/uc/container-ocr-2.jpg (example boxes: Code, Container)
- Photo 3: https://aivisionbot.com/img/uc/container-ocr-3.jpg (example boxes: Code, Container)
- Photo 4: https://aivisionbot.com/img/uc/container-ocr-4.jpg (example boxes: Container)
- 3D demo: https://aivisionbot.com/#uc-container-ocr

### 7. Crack Detection (Manufacturing and Logistics)
- Detects: Crack
- Detect: crack
- Recognize: Separates real cracks from joints and stains.
- Analyze: Measures crack length and width from calibrated images.
- Process: Ranks defects by severity and location.
- Inform: Builds a repair list with photos and coordinates.
- Engines: VYROX PixelBlade
- Value: Repairs planned by evidence.
- Photo 1: https://aivisionbot.com/img/uc/crack.jpg (example boxes: Crack, Drone)
- Photo 2: https://aivisionbot.com/img/uc/crack-2.jpg (example boxes: Crack)
- Photo 3: https://aivisionbot.com/img/uc/crack-3.jpg (example boxes: Crack)
- Photo 4: https://aivisionbot.com/img/uc/crack-4.jpg (example boxes: Crack)
- 3D demo: https://aivisionbot.com/#uc-crack

### 8. Water Meter Reading (Manufacturing and Logistics)
- Detects: Meter, Digits, Dial
- Detect: meter
- Recognize: Reads the digits and the pointer dials.
- Analyze: Combines them into one reading and checks it against the last one.
- Process: Flags readings that jump or go backwards.
- Inform: Sends the reading to billing automatically.
- Engines: VYROX HawkEye Nano, OCR
- Value: No more manual meter rounds.
- Photo 1: https://aivisionbot.com/img/uc/meter.jpg (example boxes: Digit Window, Meter Face, Sub Dial)
- 3D demo: https://aivisionbot.com/#uc-meter

### 9. Book Spine and Shelf Label Reading (Manufacturing and Logistics)
- Detects: Spine, Title
- Detect: shelf-ocr
- Recognize: Reads titles and codes, even rotated.
- Analyze: Compares the shelf with the catalogue.
- Process: Finds misplaced and missing items.
- Inform: Gives staff a fix list by shelf.
- Engines: VYROX HawkEye, VYROX ReadSight, VYROX VisionMind
- Value: Stock takes in minutes.
- Photo 1: https://aivisionbot.com/img/uc/shelf-ocr.jpg (example boxes: Spine Read, Staff, Tablet Scanner)
- 3D demo: https://aivisionbot.com/#uc-shelf-ocr

### 10. Pig Detection, Tracking and Counting (Agriculture and Aquaculture)
- Detects: Pig, Gate Line
- Detect: pig-count
- Recognize: Tracks each pig with an ID through crowding.
- Analyze: Counts pigs in and out across the gate line.
- Process: Reconciles counts with the movement record.
- Inform: Alerts on count mismatches.
- Engines: VYROX HawkEye, VYROX TrailLock
- Value: Accurate headcounts at every move.
- Photo 1: https://aivisionbot.com/img/uc/pig-count.jpg (example boxes: Camera, Pig)
- 3D demo: https://aivisionbot.com/#uc-pig-count

### 11. Pig Farrowing Detection (Agriculture and Aquaculture)
- Detects: Sow, Piglet
- Detect: farrowing
- Recognize: Recognises the onset of farrowing behaviour.
- Analyze: Times each birth and the gaps between them.
- Process: Flags long gaps that may need help.
- Inform: Calls the stockperson to the right crate.
- Engines: VYROX HawkEye, action model
- Value: Fewer losses at birth.
- Photo 1: https://aivisionbot.com/img/uc/farrowing.jpg (example boxes: Heat Lamp, Piglets x7 Nursing, Sow Lying)
- 3D demo: https://aivisionbot.com/#uc-farrowing

### 12. Piglet Squeal Detection (Agriculture and Aquaculture)
- Detects: Sow, Piglet, Squeal
- Detect: squeal
- Recognize: Separates distress from normal sounds.
- Analyze: Checks if a piglet is under the sow.
- Process: Combines sound and video into one event.
- Inform: Sends a Squeal Detected alarm with the crate number.
- Engines: Audio classifier, VYROX HawkEye Nano
- Value: Fewer piglets crushed by the sow.
- Photo 1: https://aivisionbot.com/img/uc/squeal.jpg (example boxes: Heat Lamp, Microphone, Piglet, Sow, Squealing Piglet)
- 3D demo: https://aivisionbot.com/#uc-squeal

### 13. Cow and Goat Herd Counting (Agriculture and Aquaculture)
- Detects: Cow, Goat
- Detect: herd
- Recognize: Tells cows from goats.
- Analyze: Counts each species across the field.
- Process: Compares with yesterday and the farm record.
- Inform: Reports missing animals and where the herd is.
- Engines: VYROX HawkEye Pro, drone imagery
- Value: Headcounts without a muster.
- Photo 1: https://aivisionbot.com/img/uc/herd.jpg (example boxes: Cow, Goat)
- 3D demo: https://aivisionbot.com/#uc-herd

### 14. Poultry and Geese Counting (Agriculture and Aquaculture)
- Detects: Goose, Count Line
- Detect: geese
- Recognize: Tracks the flock as it crosses the line.
- Analyze: Counts the flock in real time.
- Process: Checks the count against the batch size.
- Inform: Logs the count with a time stamp.
- Engines: VYROX HawkEye Nano, density model
- Value: Fast, accurate flock counts.
- Photo 1: https://aivisionbot.com/img/uc/geese.jpg (example boxes: Farm Worker, Goose)
- 3D demo: https://aivisionbot.com/#uc-geese

### 15. Tomato Detection and Ripeness (Agriculture and Aquaculture)
- Detects: Ripe, Turning, Green
- Detect: tomato
- Recognize: Classifies ripeness from green to red.
- Analyze: Counts ripe fruit per row and forecasts harvest.
- Process: Plans picking routes for the crew or robot.
- Inform: Shows ready to pick fruit per row.
- Engines: VYROX HawkEye, classifier
- Value: Pick at the right day, less waste.
- Photo 1: https://aivisionbot.com/img/uc/tomato.jpg (example boxes: Harvest Robot, Ripe, Turning)
- 3D demo: https://aivisionbot.com/#uc-tomato

### 16. Shrimp Post-Larvae Counting (Agriculture and Aquaculture)
- Detects: Larva
- Detect: larvae
- Recognize: Separates larvae from debris and bubbles.
- Analyze: Counts the tray and estimates size.
- Process: Works out stocking density for each pond.
- Inform: Prints the count on the sale record.
- Engines: Density model, VYROX HawkEye Nano
- Value: Fair counts for buyer and seller.
- Photo 1: https://aivisionbot.com/img/uc/larvae.jpg (example boxes: Counting Area)
- 3D demo: https://aivisionbot.com/#uc-larvae

### 17. Fish Detection and Tracking (Agriculture and Aquaculture)
- Detects: Fish
- Detect: koi
- Recognize: Tracks fish with IDs as they swim.
- Analyze: Counts fish and watches feeding activity.
- Process: Spots fish that stop moving or swim oddly.
- Inform: Alerts staff to check the pond.
- Engines: VYROX HawkEye, VYROX TrailLock
- Value: Earlier signs of sickness.
- Photo 1: https://aivisionbot.com/img/uc/koi.jpg (example boxes: Koi)
- 3D demo: https://aivisionbot.com/#uc-koi

### 18. Construction Site PPE Detection (Safety and Health)
- Detects: Helmet, Vest, No Helmet, Person
- Detect: ppe
- Recognize: Matches PPE to each person.
- Analyze: Finds anyone without a helmet or vest.
- Process: Ignores brief misses with a short time filter.
- Inform: Sends a PPE alert with a snapshot to the supervisor.
- Engines: VYROX HawkEye, VYROX TrailLock
- Value: Fewer injuries, audit ready proof.
- Photo 1: https://aivisionbot.com/img/uc/ppe.jpg (example boxes: Helmet, No Helmet, Worker)
- Photo 2: https://aivisionbot.com/img/uc/ppe-2.jpg (example boxes: Helmet, No Helmet, Person, Vest)
- 3D demo: https://aivisionbot.com/#uc-ppe

### 19. Fall Detection and Alert (Safety and Health)
- Detects: Person, Fallen
- Detect: fall
- Recognize: Recognises a fall from posture and speed.
- Analyze: Checks the person stays down beyond a few seconds.
- Process: Confirms the event and blurs the snapshot.
- Inform: Calls family or staff at once.
- Engines: VYROX MotionIQ, action model
- Value: Help arrives in seconds.
- Photo 1: https://aivisionbot.com/img/uc/fall.jpg (example boxes: Chair Tipped, Fallen Person)
- 3D demo: https://aivisionbot.com/#uc-fall

### 20. Collapse and Inactivity Detection (Safety and Health)
- Detects: Person, Collapsed
- Detect: stroke
- Recognize: Recognises a sudden collapse.
- Analyze: Measures how long the person is not moving.
- Process: Raises the priority with time.
- Inform: Alerts the caregiver with the location.
- Engines: VYROX MotionIQ, action model
- Value: Faster help when every minute counts. Not a medical device.
- Photo 1: https://aivisionbot.com/img/uc/stroke.jpg (example boxes: Collapsed Person)
- 3D demo: https://aivisionbot.com/#uc-stroke

### 21. Drowning Detection (Safety and Health)
- Detects: Swimmer, Lost
- Detect: drowning
- Recognize: Tracks each one and their movement pattern.
- Analyze: Finds a swimmer who stops or goes under.
- Process: Counts seconds underwater for that ID.
- Inform: Sounds a Swimmer Lost alarm to the lifeguard.
- Engines: VYROX HawkEye, tracker
- Value: Helps lifeguards respond sooner; never replaces supervision.
- Photo 1: https://aivisionbot.com/img/uc/drowning.jpg (example boxes: Swimmer Still 6s)
- Photo 2: https://aivisionbot.com/img/uc/drowning-2.jpg (example boxes: Lost, Swimmer)
- Photo 3: https://aivisionbot.com/img/uc/drowning-3.jpg (example boxes: Lost, Swimmer)
- 3D demo: https://aivisionbot.com/#uc-drowning

### 22. Riverbank Child Safety Zone (Safety and Health)
- Detects: Child, Adult, Danger Zone
- Detect: riverbank
- Recognize: Tells children from adults by size and pose.
- Analyze: Checks if a child enters the danger zone alone.
- Process: Raises the level the closer they get.
- Inform: Plays a warning and alerts the guard.
- Engines: VYROX HawkEye, zone rules
- Value: Prevention before an accident.
- Photo 1: https://aivisionbot.com/img/uc/riverbank.jpg (example boxes: Camera, Child, Danger Zone, Person)
- 3D demo: https://aivisionbot.com/#uc-riverbank

### 23. Smoking Detection (Safety and Health)
- Detects: Person, Cigarette, Smoke
- Detect: smoking
- Recognize: Recognises a cigarette and smoke.
- Analyze: Confirms it over several frames.
- Process: Counts occurrences per area.
- Inform: Alerts site safety for no smoking zones.
- Engines: VYROX HawkEye, small object model
- Value: Fire risk areas stay smoke free.
- Photo 1: https://aivisionbot.com/img/uc/smoking.jpg (example boxes: Camera, Cigarette, No Smoking Sign, Person, Smoke)
- 3D demo: https://aivisionbot.com/#uc-smoking

### 24. Baby Cry and Activity Detection (Safety and Health)
- Detects: Baby, Cry
- Detect: baby
- Recognize: Recognises crying versus normal sounds.
- Analyze: Checks movement and position.
- Process: Combines sound and video into one state.
- Inform: Notifies parents on their phone.
- Engines: Audio classifier, pose
- Value: An extra pair of eyes; never replaces supervision.
- Photo 1: https://aivisionbot.com/img/uc/baby.jpg (example boxes: Baby, Baby Monitor)
- Photo 2: https://aivisionbot.com/img/uc/baby-2.jpg (example boxes: Baby, Cry)
- Photo 3: https://aivisionbot.com/img/uc/baby-3.jpg (example boxes: Baby, Cry)
- Photo 4: https://aivisionbot.com/img/uc/baby-4.jpg (example boxes: Baby, Cry)
- 3D demo: https://aivisionbot.com/#uc-baby

### 25. Driver Drowsiness Detection (Safety and Health)
- Detects: Face, Eyes Closed, Yawn
- Detect: drowsy
- Recognize: Measures eye closure and yawning.
- Analyze: Computes PERCLOS and head nodding.
- Process: Detects microsleep within a second.
- Inform: Sounds an in cab alarm and alerts the fleet office.
- Engines: VYROX MotionIQ Face
- Value: Fewer fatigue crashes.
- Photo 1: https://aivisionbot.com/img/uc/drowsy.jpg (example boxes: DMS Camera, Driver, Eyes Closed, Face, Vehicle Ahead)
- Photo 2: https://aivisionbot.com/img/uc/drowsy-2.jpg (example boxes: Eyes Closed, Face, Yawn)
- 3D demo: https://aivisionbot.com/#uc-drowsy

### 26. Indoor Fire and Flame Detection (Fire and Environment)
- Detects: Fire, Smoke
- Detect: fire-indoor
- Recognize: Separates real fire from lights and reflections.
- Analyze: Confirms growth over several seconds.
- Process: Raises the level as it spreads.
- Inform: Can alert staff early, alongside certified detectors.
- Engines: VYROX HawkEye, temporal filter
- Value: Earlier response alongside certified alarms.
- Photo 1: https://aivisionbot.com/img/uc/fire-indoor.jpg (example boxes: Camera, Fire, Staff)
- 3D demo: https://aivisionbot.com/#uc-fire-indoor

### 27. Outdoor Early Fire Warning (Fire and Environment)
- Detects: Fire, Smoke
- Detect: fire-outdoor
- Recognize: Filters sun glare and steam.
- Analyze: Tracks the fire size over time.
- Process: Marks the location on the site map.
- Inform: Notifies the patrol team.
- Engines: VYROX HawkEye, temporal filter
- Value: Small fires found before they spread.
- Photo 1: https://aivisionbot.com/img/uc/fire-outdoor.jpg (example boxes: Camera, Fire, Person, Smoke)
- 3D demo: https://aivisionbot.com/#uc-fire-outdoor

### 28. Early Wildfire Smoke Detection (Fire and Environment)
- Detects: Smoke
- Detect: wildfire
- Recognize: Separates smoke from cloud and haze.
- Analyze: Confirms rising, growing smoke.
- Process: Estimates the bearing from the tower.
- Inform: Alerts fire services with the location.
- Engines: Smoke model, temporal filter
- Value: Minutes saved at the start of a fire.
- Photo 1: https://aivisionbot.com/img/uc/wildfire.jpg (example boxes: Fire Watch Camera, Smoke)
- 3D demo: https://aivisionbot.com/#uc-wildfire

### 29. Water Level Monitoring (Fire and Environment)
- Detects: Gauge, Water Line
- Detect: water-level
- Recognize: Reads the scale markings.
- Analyze: Converts the line to a level in metres.
- Process: Tracks the rate of rise.
- Inform: Warns when levels pass the alert line.
- Engines: Segmentation, OCR
- Value: Flood warnings from any camera.
- Photo 1: https://aivisionbot.com/img/uc/water-level.jpg (example boxes: Camera, Gauge, Water Line 41.52 m)
- 3D demo: https://aivisionbot.com/#uc-water-level

### 30. Thermal Human Detection (Fire and Environment)
- Detects: Person
- Detect: thermal
- Recognize: Separates people from animals and warm objects.
- Analyze: Tracks movement across the area.
- Process: Checks against restricted zones.
- Inform: Alerts the control room.
- Engines: VYROX HawkEye Thermal
- Value: See people where normal cameras cannot.
- Photo 1: https://aivisionbot.com/img/uc/thermal.jpg (example boxes: Person)
- 3D demo: https://aivisionbot.com/#uc-thermal

### 31. Multi-Lane Vehicle Counting and Speed (Mobility and Airspace)
- Detects: Car, Bus, Truck
- Detect: vehicle-count
- Recognize: Classifies car, bus and truck.
- Analyze: Counts per lane and estimates speed from tracks.
- Process: Builds hourly volumes and flags speeding.
- Inform: Feeds dashboards and traffic reports.
- Engines: VYROX HawkEye, VYROX TrailLock
- Value: Traffic data from existing cameras.
- Photo 1: https://aivisionbot.com/img/uc/vehicle-count.jpg (example boxes: Bus, Car, Truck)
- 3D demo: https://aivisionbot.com/#uc-vehicle-count

### 32. Red-Light Running Detection (Mobility and Airspace)
- Detects: Motorcycle, Car, Stop Line
- Detect: red-light
- Recognize: Tracks each vehicle approaching the line.
- Analyze: Finds vehicles crossing while the light is red.
- Process: Records a clip as evidence.
- Inform: Sends the case for review.
- Engines: VYROX HawkEye, tracker, light state
- Value: Safer junctions.
- Photo 1: https://aivisionbot.com/img/uc/red-light.jpg (example boxes: Car, Motorcycle, Red Light, Truck)
- 3D demo: https://aivisionbot.com/#uc-red-light

### 33. Parking Lot Occupancy (Mobility and Airspace)
- Detects: Car, Free Bay
- Detect: parking
- Recognize: Matches cars to painted bays.
- Analyze: Counts free and taken bays.
- Process: Updates occupancy every few seconds.
- Inform: Drives signs and an app map.
- Engines: VYROX HawkEye, bay map
- Value: Drivers find a bay faster.
- Photo 1: https://aivisionbot.com/img/uc/parking.jpg
- 3D demo: https://aivisionbot.com/#uc-parking

### 34. UAV and Drone Detection (Mobility and Airspace)
- Detects: Drone, Bird
- Detect: uav
- Recognize: Tells drones from birds.
- Analyze: Tracks each drone path.
- Process: Checks against protected airspace.
- Inform: Alerts security with the bearing.
- Engines: VYROX HawkEye Pro, tracker
- Value: Protected sites stay protected.
- Photo 1: https://aivisionbot.com/img/uc/uav.jpg (example boxes: Drone)
- 3D demo: https://aivisionbot.com/#uc-uav

### 35. Restaurant Table Occupancy (Retail and Spaces)
- Detects: Table Free, Table Busy, Person
- Detect: tables
- Recognize: Knows which tables are free, seated or waiting.
- Analyze: Measures dwell time per table.
- Process: Spots tables waiting too long for service.
- Inform: Tells staff which table to clear or serve.
- Engines: VYROX HawkEye, zone rules
- Value: Faster table turns.
- Photo 1: https://aivisionbot.com/img/uc/tables.jpg (example boxes: Guest, Table)
- 3D demo: https://aivisionbot.com/#uc-tables

### 36. Shopper Heatmap and Tracking (Retail and Spaces)
- Detects: Shopper
- Detect: heatmap
- Recognize: Tracks paths through the store.
- Analyze: Builds a heatmap of time spent per zone.
- Process: Compares layouts and promotions.
- Inform: Shows hot and cold zones daily.
- Engines: VYROX HawkEye, VYROX TrailLock
- Value: Better layouts from real behaviour.
- Photo 1: https://aivisionbot.com/img/uc/heatmap.jpg (example boxes: Shopper)
- 3D demo: https://aivisionbot.com/#uc-heatmap

### 37. Hand Gesture Control (Retail and Spaces)
- Detects: Person, Hand, Open Palm
- Detect: gesture
- Recognize: Recognises gestures like open palm, one and two.
- Analyze: Confirms the gesture for half a second.
- Process: Maps it to a command.
- Inform: Turns on the light or runs the action.
- Engines: VYROX MotionIQ Hands
- Value: Touch free control.
- Photo 1: https://aivisionbot.com/img/uc/gesture.jpg (example boxes: Gesture Camera, Open Palm, Person, Smart Light)
- 3D demo: https://aivisionbot.com/#uc-gesture

### 38. Rope Skipping and Exercise Counting (Retail and Spaces)
- Detects: Person, Jump
- Detect: rope
- Recognize: Recognises each jump from hip height.
- Analyze: Counts jumps and the rhythm.
- Process: Scores form and consistency.
- Inform: Shows the count and a summary.
- Engines: VYROX MotionIQ
- Value: Coaching from a phone camera.
- Photo 1: https://aivisionbot.com/img/uc/rope.jpg (example boxes: Person, Rope, Tablet Camera)
- 3D demo: https://aivisionbot.com/#uc-rope

### 39. Weapon Detection (Security)
- Detects: Person, Pistol
- Detect: weapon
- Recognize: Recognises a visible firearm.
- Analyze: Confirms over several frames to avoid false alarms.
- Process: Locks onto the person track.
- Inform: Alerts the control room with a snapshot for a person to verify.
- Engines: VYROX HawkEye, verifier
- Value: An aid to human monitoring, not a security guarantee.
- Photo 1: https://aivisionbot.com/img/uc/weapon.jpg (example boxes: Camera, Hand Concealed, Motorcycle, Person)
- 3D demo: https://aivisionbot.com/#uc-weapon

## FAQ

### What Is AiVisionBot?
A computer vision service from VYROX AI in Malaysia. We pick the use case with you, train on your own footage, install on site and stay on to tune it, for factories, farms, quarries, roads, shops and safety teams.

### Do I Need New Cameras?
Usually not. Most IP and CCTV cameras at 1080p work. Small objects, fine defects, underwater views and long range need the right lens, light and mounting, which we check on site.

### Does Video Leave My Site?
It does not have to. Models can run on an edge box on site and send only events, snapshots and counts. Cloud processing is optional.

### Are the 3D Demos Real Model Output?
No. The 3D demos are simulations made to explain each use case. On a real project we train and test models on your own footage and share the accuracy before you commit.

### Which AI Models Do You Use?
Our own VYROX Vision Engine. Each use case uses only the engines it needs, and each is retrained on footage from your site. The engine table above shows what each one does.

### How Do I Start?
Send a 30 to 60 second clip from the camera you would use, taken at a normal working time, and one sentence on the problem. A phone recording of the screen is fine to begin with.