Background

This case study focuses on the Media Search Engine project delivered by WEBSENSA for their client, WFDiF (Documentary and Feature Film Studios). AXII, a sophisticated tool created by Impeccable.AI, played a key role in this project.

AXII is a platform designed for creating AI-based solutions, that streamlines team collaboration and optimises resources. It centralises knowledge and simplifies infrastructure management, allowing teams to focus on innovations rather than the process.

Understanding the Client

WFDiF (Documentary and Feature Film Studios) is a cornerstone of Polish cinema with a rich 70-year history in film production. Additionally, WFDiF serves as a vast repository of numerous archival and contemporary films.

WFDiF sought a solution enabling easy search capabilities throughout their extensive collection of audiovisual materials. Their existing system relied mainly on manually entered metadata, which proved inadequate in the process of extracting information from these materials.

Understanding the Project

The goal of the Media Search Engine project was to enhance the search functions of the 35mm.online portal by enabling the search through audiovisual materials from the WFDiF's library using artificial intelligence.

Using modern AI models, WEBSENSA created the solution that autonomously identifies a broad range of elements in WFDiF's materials, such as various objects, architecture, music pieces, actors, celebrities, scene locations, emotions, words, and sounds.

Integration of the AXII Platform

The AXII played a key role in the Media Search Engine project by enhancing collaboration between data science and engineering teams, and accelerating the integration of core AI functionalities into the production service. This facilitated a more efficient workflow and expedited the development process, ultimately contributing to the project's success.

The simplicity of integration can be attributed to the following factors:

1. Intuitive web browser client: AXII's clear and visually appealing user interface enabled seamless navigation and interaction, making it easy for developers to understand and utilise the platform's functionalities.
2. Easy setup: AXII could be added to the current environment without the extra frameworks or additional dependencies, allowing for a hassle-free integration.
3. Risk-free adoption: With no vendor lock-in or long-term contracts, AXII offered flexibility and could be easily incorporated into the tech stack without hidden risks.
4. Compatibility with existing tools: Designed to work seamlessly with popular frameworks and libraries, AXII could be effortlessly added to the current stack, enabling users to immediately see its results.

These factors ensured a smooth and straightforward integration of the AXII platform into the Media Search project, allowing the team to fully leverage its capabilities and achieve impressive results.

Role of the AXII in Project Implementation

AXII's integration helped WEBSENSA to face the challenge of applying modern AI models to classify elements in WFDiF's archival (often black-and-white) audiovisual materials. The platform enabled the adaptation of state-of-the-art solutions to datasets not previously analysed in this way.

Key features of the AXII platform included:

• Launching complex AI workflows – a simplified process of initialising and orchestrating complex AI workflows,
• Automatic resources management – automated management of computational resources and data, leading to significant cost savings,
• Workflow visualisation – a comprehensive visualisation of the entire film processing process, enhancing understandability for the whole team,
• Knowledge base – a centralised knowledge repository that safeguards project know-how, maintaining important knowledge over the project's 5-year duration,
• Flexibility – a flexible architecture allowing easy adaptations to evolving needs.

Results

The Media Search Engine project supported by the AXII platform achieved impressive results, including:

• the analysis of 36,280 minutes of films and chronicles,
• the detection of 72,504 objects in the audiovisual materials.

The AXII platform significantly impacted this project by organising and simplifying the process:

• The film processing workflow consisted of 67 steps, each contributing to various tasks such as data processing and transformation or AI model analysis.
• The workflow execution optimisation feature in AXII prevented repeating previously performed calculations, which saved:

• approximately 1,700 hours = 70 days of GPU operation time,
• 71 days that the data science team would lose waiting for calculations to complete.

AXII's substantial efficiency and time-saving features supported the successful completion of the Media Search Engine project.

Conclusions and Future Plans

Through this project, WEBSENSA and Impeccable.AI gained unique expertise in analysing audiovisual materials of varying quality, age, and colour registration.

The companies plan to leverage this newfound knowledge in future projects, as they can now easily analyse a wide range of audiovisual materials.

Quotes and Opinions

About Media Search Engine:

"Our unique achievement in this project was a quick development of a solution recognising Polish actors in films with high accuracy. There may not be an alternative solution like this on the market." — Szymon Drejewicz, Project Manager from WEBSENSA.

About AXII:

"AXII's functionalities were easy to understand. Thanks to the platform, the integration of backend with AI modules was significantly simplified. It allowed our data scientists to cooperate with developers in creating production-level workflows." — Rafał Klusek, Backend Developer from WEBSENSA

"The integration of AXII simplified our system architecture. AXII facilitates the complex operations of the AI module, making it effortless and straightforward for developers." — Bartek Starzec, Backend Developer from WEBSENSA

‍