.png?width=500&height=278&name=Customer%20Cases%20image%20(20).png)
Digital Asset Management Transformation
Optimizing the management, organization, and accessibility of digital assets like images and videos.
Data asset management system modernisation to manage assets, improve search and implement ML for better organization and recommendations.
Assets
predictable URLs
redirection
The Goal
To create a more intuitive, efficient, and accessible DAM system using advanced cloud-native solutions on Google Cloud Platform (GCP).
The Challenge
The customer faced significant challenges with its existing Digital Asset Management (DAM) system. Managing over 25,000 digital assets relied heavily on manual file naming, which led to inconsistencies, errors, and inefficiencies that ultimately impacted workflows and productivity. The customer set out to modernize this system to improve efficiency, reduce human error, and enhance scalability.
About our customer
A global brand that specializes in luxury home and body care products, it has expanded internationally over the years and has stores in several countries. It offers both physical products and online sales, aiming to help customers enhance their overall sense of wellness.
Start Your Success Story with Crystalloids
Partner with Crystalloids and Google Cloud to bring your project to life.
The Solution
Predictable URLs for Improved Asset Access
We implemented predictable URLs to improve the accessibility and consistency of digital assets. Traditionally, URLs were based on filenames, introducing risks like inconsistent naming, typos, or capitalization errors. We addressed this challenge by using asset tags within Adobe Experience Manager (AEM) to create virtual URLs that are structured, SEO-friendly, and consistently tagged. This provided predictable and reliable access for internal teams.
To date, over 250,000 virtual URLs have been created, with each asset being accessible via approximately 10 predictable URLs. The system was designed to handle URL redirection in under 50 milliseconds, ensuring rapid access for users.
Enhancing Similarity Detection with Machine Learning
This improvement replaced manual file management with a structured, automated process, significantly reducing time spent troubleshooting file-naming issues and improving the user experience for internal teams.
The DAM transformation also introduced new methods to enhance similarity detection, which helped reduce redundant or duplicate content in the asset collection. Initially, perceptual hashing was used to calculate visual fingerprints of images, effectively identifying duplicates even when properties like format, size, or color varied.
Building on this, we transitioned to machine learning-based similarity detection using Google’s BigQuery Vector Index. By utilizing machine learning models, we calculated feature vectors for each image, capturing the visual essence of the assets. These feature vectors were indexed in BigQuery's Vector Index, enabling efficient similarity searches across the entire asset library.
This ML-based approach captured subtle asset variations, such as texture, colour patterns, and context. It allowed the DAM system to suggest related assets, helping creative and marketing teams quickly find relevant images.
Additionally, similarity detection enabled the development of a recommendation engine to suggest related products based on visual similarities, improving the customer experience.
Elastic Search Integration for Enhanced Queries
We integrated Elastic Search within the client's GCP infrastructure to index all asset metadata, replacing traditional RESTful services with a more powerful query system. This improved asset search speed, flexibility, and efficiency using GCP's cloud-native capabilities.
Scalable Cloud Architecture for High Performance
The new DAM system used GCP services to focus on scalability and performance. Google Cloud Run was used for deploying containerized APIs, including the Predictable URL API, ensuring rapid scaling and integration of cloud-native functions.
The event-driven architecture powered by Google Cloud Pub/Sub allowed the automatic processing of assets when new digital files were added or updated, ensuring seamless and efficient management.
High-Level Design Overview
The high-level design of the DAM system focused on creating a modular, scalable, and resilient architecture:
- Google Cloud Datastore and BigQuery: Used to store metadata and handle large datasets for fast retrieval and scalable storage.
- Google Cloud Run: Deployed containerized APIs for predictable URLs, ensuring rapid scaling.
- BigQuery Vector Index: Enabled advanced machine learning-based similarity detection across the asset library.
- Elastic Search on GCP: Provided a powerful querying interface for asset metadata, making searches fast and flexible.
- Perceptual Hash Calculation: Used to detect asset similarities, triggering updates in metadata stored in BigQuery and Cloud Datastore.
The event-driven processing pipeline ensured that each asset update, deletion, or creation triggered automated steps, such as metadata updates, perceptual hash calculations, and feature extraction. This streamlined approach ensured that assets were uniformly processed and accessible via consistent, predictable URLs.
The Result
Through these innovations, we transformed the client's Digital Asset Management system. Predictable URLs provided consistency and ease of access, machine learning-driven similarity detection reduced asset duplication and enabled smarter recommendations, and Elastic Search enhanced asset search performance.
We helped the customer successfully modernise its DAM infrastructure by leveraging GCP cloud-native design patterns, machine learning techniques, and modular architecture.
This transformation supports creativity and innovation while reducing time spent managing inconsistencies, demonstrating comprehensive application development and modernization capabilities to the Google Cloud team.
