AI-Based Product Counting Solution for a FMCG Company
A major FMCG enterprise engaged with ThirdEye Data to resolve recurring losses during product dispatches. Relying heavily on third-party logistics for delivery operations across vast geographies, the company faced significant discrepancies between dispatched and received product quantities—leading to operational inefficiencies and financial losses. To solve this, ThirdEye developed a scalable AI-powered product counting solution using computer vision technology to detect and count SKUs from images and videos captured during loading and unloading. The system, now in its Proof of Concept (PoC) phase, brings new levels of accuracy, transparency, and automation to the company’s logistics processes.
THE CUSTOMER
BUSINESS GOALS OR CHALLENGES
Business Goals
- Enhance accuracy of product counts during dispatch and delivery.
- Create a reliable and verifiable audit trail for every dispatch event.
- Minimize financial losses due to quantity mismatches.
- Replace manual counting with automated, scalable computer vision solutions.
- Deploy a standardized system across multiple logistics partners and distribution centers.
Understanding the Challenges:
- Frequent mismatches between dispatched and received product quantities.
- Lack of visibility during the loading and unloading process.
- Inability to trace discrepancies due to absence of visual evidence.
- Manual counting was error-prone, inefficient, and unscalable across 200+ SKUs.
- Need to support open and closed truck environments with varied camera angles and lighting conditions.
Prerequisites and Preconditions:
- Visual input (images or videos) captured during loading and unloading of goods.
- Support for Android and iOS mobile platforms for wide accessibility.
- Secure login authentication for authorized personnel.
- AI models trained to recognize diverse SKUs across packaging formats.
- Infrastructure to host the backend AI engine and analytics dashboard.
THE SOLUTION
To address the ongoing challenges of dispatch discrepancies and lack of visibility, an AI-powered product counting solution was developed and deployed as a mobile-first Proof of Concept (PoC). This computer vision–driven system allows users to capture or upload images and videos during product loading and unloading. Using deep learning models trained on the company’s diverse SKU set, the solution automatically detects and counts products in real time. A built-in quality check ensures visual inputs meet clarity and visibility standards before processing. The solution provides an intuitive interface, seamless backend processing, and a centralized analytics dashboard, enabling greater transparency, accountability, and operational efficiency. Designed to be scalable and adaptable, it lays the foundation for deployment across multiple logistics partners and geographies.
Solution Highlights
-
Developed a mobile-first AI-powered application for product counting using computer vision.
-
Integrated secure login and authentication for authorized users.
-
Enabled users to upload or capture real-time images/videos of goods during dispatch.
-
Implemented automated image quality checks for clarity, lighting, and product visibility.
-
Backend AI engine detects and counts products from approved visuals.
-
Real-time count displayed on user device, validating loading and unloading events.
-
Analytics dashboard provides insights into total counts, user performance, and process metrics.
-
Built-in feedback mechanism allows users to report issues and suggest improvements.
-
Scalable architecture designed to adapt to new SKUs, logistics setups, and geographies.
-
MVP version under development for production deployment and multi-center integration.
Technologies Used
-
Computer Vision & AI Models
-
YOLOv8 for object detection and SKU counting
-
PyTorch for model training and inference
-
Roboflow for dataset annotation and augmentation
-
-
Mobile Application & UI
-
Android & iOS support with unified codebase
-
Streamlined, user-friendly interface for guided image capture
-
Secure login and session handling
-
-
Backend Infrastructure
-
FastAPI for scalable API services
-
Docker for containerized deployment and portability
-
PostgreSQL for data storage and session logging
-
-
Analytics & Visualization
-
Integrated dashboard for performance monitoring
-
Aggregated insights per user, session, and location
-
Exportable reports for audits and process review
-
-
Security & Data Privacy
-
End-to-end data encryption during transmission and storage
-
Authentication and role-based access control
-
Compliance with data privacy and operational security standards
-
VALUE CREATED
- 80–90% accuracy achieved in product counting across diverse packaging types and truck configurations
- Eliminated manual counting at pilot locations, saving over 50% of operational time during dispatch
- Visual audit trail enables traceability and accountability for discrepancies
- Loss prevention with early detection of mismatches between loading and delivery
- Real-time transparency for operations teams across geographies
- High scalability: the solution architecture supports expansion to multiple distribution centers and 3PL sites
- Reduced financial risk associated with unverified missing product claims
- Enhanced trust and operational alignment between internal teams and third-party partners
Transforming Enterprises with
Data & AI Services & Solutions.
ThirdEye delivers Data and AI services & solutions for enterprises worldwide by
leveraging state-of-the-art Data & AI technologies.
