Accelerating Computer Vision Workflows with Databricks
Intuz Development & Consulting
- Data Collection & Preprocessing
- Model Development
- Real-time Processing and Scalability
- Integration with Business Systems
About the Project
A leading retail chain with a strong online presence and physical stores globally wanted to enhance its inventory management and customer experience through real-time shelf monitoring. The company faced challenges in tracking stock levels, misplaced items, and detecting damaged products across its warehouses and retail stores.
To overcome these challenges, we developed computer vision solution using artificial intelligence. The goal was to automate product tracking, minimize stockouts, and optimize inventory management. The solution was built on the Databricks platform to handle large-scale image data processing and analytics in real time.
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.
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
- The camera were placed to capture a high-resolution images were captured using in-store cameras.
- Data pipelines were built to ingest and preprocess the image data using Apache Spark framework on Databricks.
- Image augmentation techniques were improvised and new image were generated to improve model performance.
Model Development
- Convolutional Neural Networks (CNNs) based solution were developed to detect products, misplaced items, and empty shelf spaces.
- Transfer learning with pre-trained models (ResNet, EfficientNet) was used for faster deployment.
- MLflow was also used to keep the track of newly models train and do the versioning of the models.
Real-time Processing and Scalability
- The solution utilized Databricks Delta Lake for optimized storage and real-time query performance.
- Auto-scaling clusters were implemented which ensured cost-effective and high-speed computations to spin up the clusters as required.
- Model inference was integrated into the Databricks ML runtime to enable real-time analysis.
Integration with Business Systems
- API endpoints were developed to integrate insights.
- A dashboard powered by Databricks SQL, CNN model inference and Power BI provided actionable insights.
Business Impact
- Improved Inventory Accuracy – Real-time stock updates reduced inventory discrepancies by 85%.
- Cost Savings – Automation reduced labour costs by 40% and improved operational efficiency.
- Enhanced Customer Satisfaction – Timely restocking led to higher customer satisfaction scores.
- Faster Decision-Making – The business gained real-time visibility into inventory trends and optimized supply chain management.
- Scalability & Future Expansion – The solution provided a scalable framework for expansion across more stores and regions.
Technical Specifications
Databricks on AWS
Explore More Work
We changed the way they do business, and they have no complaints
Marengo Load Board
Logistic Aggregator Mobile and Web App connecting Shippers, Carriers and Truck Drivers Targeting African Region
Bubba Booking
Intuitive Travel Application for Booking Last-Minute Tours & Adventure Deals
Let’s Talk
Let us know if there’s an opportunity for us to build something awesome together.