Inventray – Autonomous Inventory Intelligence

93.2%

YOLOv8 Detection Accuracy

<140ms

Processing per Image

0

Special Hardware Required

ERP

SAP & Dynamics Ready

Project Overview

Inventray is an AI-powered inventory platform that eliminates manual stock counting. Using any modern smartphone or tablet, Inventray One scans a shelf or warehouse, identifies every product using computer vision, estimates depth and spatial structure in 3D, and delivers structured inventory data ready to be pushed directly into any ERP system — in under 140ms per image.

The Problem It Solves

Manual counting is slow and error-prone
Barcode scanners require special MDE hardware
Stock errors cause lost revenue and overstock waste
Frequent recounts tie up staff and resources

Key Features

Works on any modern smartphone — no special hardware
Real-time processing under 140ms per image
Confidence scoring flags uncertain counts
3D detection for stacked/overlapping items
Full audit trail with timestamp and device ID

Product Screenshots

Scanning Interface

Results & Analytics

Analytics Dashboard

AI Detection View

AI Processing Pipeline

1

Camera Scan

Shelf or warehouse captured via smartphone

2

Object Detection

YOLOv8 identifies individual products at 93.2%

3

3D Depth

MiDaS v3 analyzes depth & spatial structure

4

3D Fusion

Intelligent 2D + 3D data merging

5

ERP Output

Structured JSON data for your system

Technical Stack

Core Technologies

Mobile: React Native / Flutter (iOS & Android)
Backend: FastAPI + PyTorch + TorchServe
AI Models: YOLOv8 (detection) + MiDaS v3 (depth)
Cloud: Docker-based, horizontally scalable
API: API-first design, ERP-agnostic JSON output

ERP Compatibility

SAP

Microsoft Dynamics

Custom / Bespoke Systems

REST API — any Warenwirtschaft backend

Industries & Use Cases

Retail

Shelf inventory in supermarkets and retail chains — fast, precise, no downtime needed during store hours.

Warehouse & Logistics

Stock capture in logistics centers and distribution warehouses — entire aisles scanned in minutes instead of hours.

Gastronomy

Ingredients, walk-in coolers, and storage rooms — reduce spoilage, improve transparency, and cut prep time.

Industry

Volume measurement for bulk goods, liquids, and irregular-shaped materials using 3D spatial understanding.

Code Sample — Inventory Event

// Example ERP Output — Structured Inventory Event (JSON)
{
  "event_type": "INVENTORY_SCAN",
  "timestamp": "2025-04-25T14:32:10Z",
  "device_id": "INV-PHONE-001",
  "location": "Shelf A3 – Cold Aisle",
  "items": [
    {
      "product_id": "SKU-98231",
      "name": "Whole Milk 1L",
      "count": 14,
      "confidence": 0.97,
      "detection_method": "YOLOv8_2D+MiDaS_3D"
    },
    {
      "product_id": "SKU-44812",
      "name": "Greek Yogurt 500g",
      "count": 8,
      "confidence": 0.88,
      "detection_method": "YOLOv8_2D+MiDaS_3D",
      "flag": "LOW_CONFIDENCE"
    }
  ],
  "scan_duration_ms": 137,
  "erp_target": "SAP_S4HANA"
}

Results from Real Test Scenarios

Measured Performance

Product Detection Accuracy 93.2%

Processing Speed <140ms/frame

ERP Integration Reliability 99.5%

Business Impact

Eliminates manual counting labor costs entirely
Reduces stock errors and associated revenue losses
No hardware procurement — uses existing devices
Audit-trail compliance out of the box