AR Counter is a mobile computer vision platform designed for real-time object counting in physical environments.
The project originated from a practical inventory problem encountered in office workflows, where repetitive manual counting introduces inefficiencies and human error.
Rather than simply detecting objects, AR Counter focuses on persistent counting, ensuring that previously counted objects are not repeatedly registered as the camera moves.
| Metric | Value |
|---|---|
| Current State | Functional MVP |
| Platform | Android |
| Deployment | Debug APK Released |
| Validation | Real-world Inventory Testing |
| Version | Early Stage |
- Persistent Object Tracking
- Counting Stability
- Low-Light Performance
- Spatial Anchoring
- User Experience Optimization
AR Counter derives counting reliability through three complementary mechanisms.
| Component | Objective |
|---|---|
| Object Detection | Identify physical entities in real time |
| Spatial Anchoring | Assign persistent coordinates to detected objects |
| Adaptive Sensitivity | Dynamically adjust duplicate suppression thresholds |
Together these mechanisms create a lightweight inventory intelligence system capable of operating entirely on mobile hardware.
| Stage | Purpose |
|---|---|
| Camera Acquisition | Capture live video stream |
| Object Detection | ML Kit inference |
| Coordinate Assignment | Spatial anchoring |
| Duplicate Suppression | Persistent ID validation |
| Sensitivity Adjustment | Dynamic threshold tuning |
| Counting Engine | Increment object count |
| Visualization | Overlay detections |
| Reporting | Inventory statistics |
| Domain | Stack |
|---|---|
| Mobile Development | Kotlin |
| Computer Vision | OpenCV |
| AI Inference | Google ML Kit |
| User Interface | XML, AppCompat |
| Tracking Logic | Spatial Anchoring |
| Deployment | Android APK |
Different objects exhibit varying densities and spacing patterns.
AR Counter introduces an adjustable sensitivity mechanism allowing users to calibrate the counting radius according to the operating environment.
Optimized for:
- Bricks
- Small Parts
- Dense Inventory
- Packed Storage
Optimized for:
- Boxes
- Office Equipment
- Pallets
- Large Objects
- Low-light degradation
- Frame skipping during rapid motion
- Occlusion handling
- Large object overlap
- Code modularization
A debug APK is available in the Releases section.
As this is an unsigned debug build, Android may display a Play Protect warning.
Select:
Install Anyway
to evaluate the application.
- Computer Vision
- Mobile Development
- Spatial Intelligence
- Edge AI
- Human-Centered Design
- Inventory Automation
- Systems Engineering
The objective was never merely counting objects.
The objective was maintaining identity through space.
AR Counter explores how lightweight computer vision systems can preserve object persistence under motion and environmental variability.
Built by Ganateju
"Spatial awareness under motion."