Accelerating Computer Vision Workflows with Databricks

A large retail store improved inventory tracking with Databricks, speeding up computer vision processes, making them more accurate, and more efficient for better real-time inventory management.

A global retail brand struggled with tracking stock levels, misplaced items, and damaged products in its stores and warehouses. We built an AI-powered computer vision solution on Databricks to monitor shelves in real time, automate product tracking, reduce stockouts, and improve inventory management and overall customer experience.

Intuz Development & Consulting

Data Collection & Preprocessing

Model Development

Real-time Processing and Scalability

Integration with Business Systems

System Architecture Overview

Problem Statement

Inefficient Stock Management

The selves were taken care by the people working in the stores. Due to the inefficiency the shelves were empty sometimes which led to bad user experience as they didn’t get what they want.

High Operational Costs

Since the human were involved to take care of all the little things, it was a labour-intensive which resulted in increased costs.

Bad Customer Experience

Customers often found empty shelves or misplaced products, leading to dissatisfaction and loss of sales.

Lack of Real-time Insights

The challenge to have real-time stock visibility existed which was the major pain point of the client that we wanted to solve which can make an impact on decision-making.

Data Collection & Preprocessing

High-resolution images were captured using in-store cameras. Data pipelines on Databricks with Apache Spark were used to ingest and preprocess the images. We also applied image augmentation to generate new images and improve the model’s performance

Model Development

We built a CNN-based system to detect products, misplaced items, and empty shelves. Using transfer learning with models like ResNet and EfficientNet sped up deployment. We also used MLflow to track new model training and manage model versions.

AI-powered computer vision on Databricks transformed retail operations by capturing and analyzing large-scale image data, automating inventory monitoring, and integrating AI/ML to provide actionable insights that boosted profitability and improved customer satisfaction.