Media Search Engine for WFDiF

Enhancing Digital Media Accessibility

Background

WEBSENSA, a leading provider of AI-based solutions, partnered with WFDiF (Documentary and Feature Film Studios) to embark on a transformative project – the Media Search Engine.

The purpose was to enrich WFDiF's 35mm.online portal with AI-driven search capabilities for audiovisual materials, which initially were labelled only with manually entered metadata.

The project aimed to augment these labelling data by automatically identifying objects in films, such as architectural features, musical compositions, actors, locations, emotions, words and sounds.

Understanding the Client

WFDiF (Documentary and Feature Film Studios), a cornerstone of Polish cinema, has a 70-year history in film production and has shaped renowned filmmakers' careers. Today, it is Poland's largest film production centre, serving film professionals and emerging talents.

WFDiF required a solution to enhance its search engine capabilities with state-of-the-art technology. They expected efficient AI software that could explore the depths of audiovisual materials and extract valuable information from them. Using it, they wanted to offer their users a superior search experience.

Project Description

The Media Search Engine project used modern AI models to recognise elements in WFDiF's archival audiovisual materials (often black-and-white). The scope and ambition of this project demanded several teams of developers from multiple subcontractors. To coordinate smoothly, protocols for system components, irrespective of the developer, were defined.

A significant tool used was the AXII platform, which enabled efficient optimisation of the configuration and integration of various components into larger structures.

WEBSENSA managed the project using agile methodologies, with every requirement documented and transformed into tasks within project management tools.

Project Execution

The project consisted of several stages, which were as follows:

1. Requirement gathering and client consultations;
2. Scheduling planning;
3. Initial development of components for object recognition in audiovisual materials;
4. API design for communication between AI components and data management;
5. Implementation and testing of API;
6. Test deployments and error corrections;
7. AI model calibration and API error corrections;
8. Transition to maintenance phase (minor modifications and continuity checks).

Results

The Media Search Engine project yielded impressive results:

• 36,280 minutes of films and newsreels analysed,
• 72,504 detections (all objects combined),
• 3,493 people detected – from film metadata,
• 427 characteristic sounds identified,
• 263 expletives detected,
• 1,421 different objects identified,
• 338 architectural elements (characteristic buildings),
• 346 locations (e.g., meadows, mountain ranges, tracks).

Conclusions and Future Directions

This project imparted insightful learnings upon WEBSENSA in the analysis of audiovisual materials, leading to the unparalleled development of a solution capable of swiftly and accurately identifying Polish actors in films – a first in the industry.

It broadened our proficiency from analysing contemporary materials to mastering diverse, archival, and even monochrome films, irrespective of their quality or age.

This experience is shaping WEBSENSA’s future direction, enabling a seamless exploration of new frontiers in film analysis.

Client's feedback

• Quality: 5.0
• Schedule: 5.0
• Cost: 5.0
• Willing to Refer: 5.0

"WEBSENSA delivered a secure and well-performing solution with a low error rate, exceeding initial expectations. They were professional, knowledgeable, and solutions-oriented." — Maciej Mrozowicz, Project Manager at WFDiF

Contact us

Interested in implementing a similar solution for your business? Contact us – our experts are ready to tailor our offer to your specific needs.