Topics In Demand
Notification
New

No notification found.

Leveraging GenAI and LLMs for Accurate Video Ad Classification
Leveraging GenAI and LLMs for Accurate Video Ad Classification

January 20, 2025

6

0

Video advertising is an integral part of the digital marketing ecosystem, playing a crucial role in engaging audiences and driving brand awareness. With the exponential rise of multimedia content, the need for efficient and accurate ad classification has become more important than ever.

By harnessing the power of GenAI and LLMs, we can automatically and precisely classify video ads according to IAB standards, enhancing targeting, optimizing ad placements, and ultimately improving the overall effectiveness of digital marketing campaigns.

In this blog, we explore the ‘IAB Video Ad Classifier,’ a GenAI-based project utilizing large language models (LLMs) to accurately classify video advertisements according to Interactive Advertising Bureau (IAB) standards. This article dives into the project’s key features, workflow, and technological components. 

1. Understanding the IAB Video Ad Classifier

The IAB Video Ad Classifier is designed to identify and categorize video advertisements based on predefined criteria from the Interactive Advertising Bureau (IAB). IAB defines 28 primary (Tier 1) categories and several sub-categories (Tier 2) adding up to a total of 316 unique categories.  It aims to streamline the process of ad classification, allowing for better ad targeting, compliance, and analytics. By leveraging AI and machine learning, this classifier offers a robust solution to the growing complexity of video ads.

2. Technologies and Tools Behind the IAB Video Ad Classifier

The IAB Video Ad Classifier employs several cutting-edge technologies including Gradio, Whisper, and Hugging Face Transformers. At its core, the classifier uses AI models to analyze video­ content, extract audio, and generate text-based captions. The models, such as Mistral and Gemini, play a crucial role in text generation and classification. Additionally, other libraries, like MoviePy, Transformers and Langchain, contribute to the audio and video processing components.

3. Architecture¬ of the IAB Video Ad Classifier

The classifier’s architecture comprises various interconnected components that work together to classify video ads. It starts with video input, which is processed to extract audio and generate captions. These captions, along with audio transcriptions, are used to identify the content’s context and categorize it according to IAB standards. This multi-step process ensures that the IAB Video Ad Classifier is accurate and reliable.

Fig 1: Ad Classification Pipeline

3. Architecture¬ of the IAB Video Ad Classifier

The workflow of the IAB Video Ad Classifier is designed for efficiency and precision. Here’s an overview of the key steps:

Model Selection: The classifier allows you to choose the AI model for inference. This flexibility lets you customize the processing pipeline based on your specific requirements.

Fig 2: Model Selection UI

Video Input: The classifier takes a video as input.

Audio Extraction: The audio track is extracted and transcriptions are generated.

Image Captioning: Frames from the video are analysed to create textual descriptions.

Classification: The transcriptions and captions are fed into an AI model to classify the video according to IAB standards.

Output: The classifier returns the appropriate IAB classification, including both primary and secondary categories.

Fig 3: Classier Output

5. Challenges and Solutions

While developing the IAB Video Ad Classifier, the team faced several challenges, such as maintaining accuracy, managing large models, and ensuring quick inference times. To overcome these challenges, the team used advanced techniques like quantization, LoRA configuration, and efficient model architectures. These solutions helped optimize the classifier for speed and precision.  

6. Applications of the IAB Video Ad Classifier

The IAB Video Ad Classifier has a wide range of applications in the advertising industry. It can be used for content moderation, ad targeting, compliance checks, and analytics. By accurately classifying video ads, businesses can ensure their advertisements are displayed to the right audience and meet industry standards. This contributes to a more efficient and effective digital advertising ecosystem.

7. Results and Performance

IAB Video Ad Classifier application pipeline generates results with high accuracy and low inference times. The classifier can process multiple video formats and handle various types of advertisements.

8. Future Developments and Enhancements

The team behind the IAB Video Ad Classifier has plans for further enhancements, such as incorporating additional AI models, improving classification accuracy, and expanding the classifier’s functionality. Future developments will include better support for different languages, integration with other advertising platforms, and enhanced user interfaces for easier interaction.

Conclusion

The IAB Video Ad Classifier is a remarkable project that demonstrates the power of AI in video ad classification. By leveraging advanced technologies, the classifier provides accurate and efficient ad categorization according to IAB standards. This project has the potential to revolutionize how video ads are managed and analyzed in the digital marketing industry. As the project continues to evolve, it will undoubtedly play a significant role in shaping the future of video advertising.


That the contents of third-party articles/blogs published here on the website, and the interpretation of all information in the article/blogs such as data, maps, numbers, opinions etc. displayed in the article/blogs and views or the opinions expressed within the content are solely of the author's; and do not reflect the opinions and beliefs of NASSCOM or its affiliates in any manner. NASSCOM does not take any liability w.r.t. content in any manner and will not be liable in any manner whatsoever for any kind of liability arising out of any act, error or omission. The contents of third-party article/blogs published, are provided solely as convenience; and the presence of these articles/blogs should not, under any circumstances, be considered as an endorsement of the contents by NASSCOM in any manner; and if you chose to access these articles/blogs , you do so at your own risk.


Ignitarium

© Copyright nasscom. All Rights Reserved.