WeAreDevelopers Logo WeAreDevelopers
Sign in
For You LIVE
World Congress 2022 • Jun 15, 2022

Shoot for the moon - machine learning for automated online ad detection

Humera Minhas , Parinitha Hirehal

When computer vision proved too heavy for real-time ad blocking, engineers pivoted to DOM analysis. Discover how they achieved sub-millisecond inference in-browser using XGBoost and TensorFlow.js.

Pause
Mute Enter Fullscreen
#1 about 3 min

Using machine learning to automate online advertisement detection

Project Moonshot aims to filter intrusive internet ads using artificial intelligence and machine learning.

#2 about 2 min

Understanding the limitations of manual ad filter lists

Traditional filter lists demand significant human intervention to manually review and update rules against circumvention.

#3 about 2 min

Establishing an experimentation pipeline for ad detection models

The end-to-end process involves data collection, pre-processing, model benchmarking, and final deployment.

#4 about 3 min

Choosing HTML structures over computer vision for ad data

Analyzing HTML tags and page structure provides better performance than using heavy computer vision models.

#5 about 2 min

Generating ground truth data at scale using web crawlers

A custom crawler combined with headless Chrome and Adblock Plus creates large-scale labeled datasets.

#6 about 3 min

Pre-processing raw HTML into feature and adjacency matrices

Converting HTML into a JSON tree structure enables the generation of matrices for machine learning input.

#7 about 3 min

Overcoming challenges with unbalanced data and pipeline speed

Data augmentation resolves highly unbalanced distributions while server upgrades drastically reduce processing times.

#8 about 5 min

Applying graph neural networks for HTML node classification

Message passing mechanisms generate node embeddings that allow classifiers to distinguish ads from legitimate content.

#9 about 5 min

Evaluating model performance and utilizing self-supervised learning

Tree-based models like XGBoost achieve better F1 scores than direct graph neural networks for complex web graphs.

#10 about 4 min

Deploying Python machine learning models into JavaScript environments

Migrating models to background scripts using TensorFlow.js reduces inference latency and prevents continuous reloading.

#11 about 5 min

Addressing model circumvention and future machine learning objectives

Continuous data collection from ad publishers attempting to bypass detection further trains and hardens the model.

Matching moments

1:33 min

Real-world implementations of browser-based machine learning

Håkan Silfvernagel · LIVE

7:51 min

Integrating embedded machine learning models into the Firefox browser

Chris Heilmann Chris Heilmann +2 · LIVE

57 sec

Implementing predictive prefetch with machine learning

Jessica Janiuk · JS Congress

2:54 min

Leveraging Chrome AI and nano models for web applications

Raymond Camden · Perf + AI

5:36 min

Evaluating page performance using customized ML prediction models

Perf + AI

10:12 min

Challenges with open source monetization and AI scraping

Chris Heilmann Chris Heilmann +2 · LIVE

More from World Congress 2022

What the heck do "declarative" and "reactive" actually mean? thumbnail
30:40
about 4 years ago

What the heck do "declarative" and "reactive" actually mean?

André Kovac

The Design and Evolution of C++ thumbnail
1:01:15
over 4 years ago

The Design and Evolution of C++

Bjarne Stroustrup

But, you're not Facebook thumbnail
31:12
over 4 years ago

But, you're not Facebook

Kristijan Ristovski

Communicate efficiently with Software Architecture Diagrams thumbnail
28:11
over 4 years ago

Communicate efficiently with Software Architecture Diagrams

Simon Lasselsberger

Get Ready for Jakarta EE 10! thumbnail
42:53
over 4 years ago

Get Ready for Jakarta EE 10!

Ivar Grimstad

Keep your code refactorable thumbnail
26:46
over 4 years ago

Keep your code refactorable

Gerrit Stapper

Upcoming sessions on this topic

Open session

Build a Zero-Cost AI Agent in Your Browser

  • Aishwarya Mathuria

    Adobe

    Computer Scientist

  • Nishant Thakur

    Adobe

    Senior Computer Scientist

Open session

React Under Fire: Rendering Performance for Streaming Agent UIs

  • Abhirupa Mitra

    Slack

    Senior Member of Technical Staff

Open session

Teaching AI Coding Agents to Build It Right the First Time

  • Sajeetharan Sinnathurai

    Microsoft

    Principal Product Manager

Open session

Agentic Forecasting with sktime: Towards Autonomous Time Series Analysis

  • Franz Kiraly

    sktime

    Director

Open session

Architecting Trusted AI: Coupling Multi-Agent Systems with dbt Semantic Store for Zero-Hallucination

  • Pushkar Mishra

    J.P. Morgan

    Senior Vice President

Open session

Beware of Strangers Bearing Code: Open Source Trust in the Agent Era

  • Vikram Vaswani

    Consultant

Related videos

Machine learning in the browser with TensorFlowjs thumbnail
38:24
LIVE • over 5 years ago

Machine learning in the browser with TensorFlowjs

Håkan Silfvernagel

Exploring the Future of Web AI with Google thumbnail
33:51
Coffee With Developers • over 1 year ago

Exploring the Future of Web AI with Google

Thomas Steiner

AI in the Open and in Browsers - Tarek Ziadé thumbnail
53:54
Coffee With Developers • 9 months ago

AI in the Open and in Browsers - Tarek Ziadé

Tarek Ziadé

Run AI models in the browser: Real life scenarios and implementations thumbnail
33:15
Europe 2026 Virtual • 4 months ago

Run AI models in the browser: Real life scenarios and implementations

Alexandru Hang

Beyond Vibe Coding: An AI Builder’s Guide to Migrating 10k+ React Components thumbnail
29:38
World Congress 2026 North America • 24 days ago

Beyond Vibe Coding: An AI Builder’s Guide to Migrating 10k+ React Components

Shubham Chaurasia

From ML to LLM: On-device AI in the Browser thumbnail
27:23
LIVE • almost 2 years ago

From ML to LLM: On-device AI in the Browser

Nico Martin

Related articles

From learning to earning

Jobs that call for the skills explored in this talk.