📹 Welcome to Brick by Brick Challenge 2024

:office: Revolutionising Building Management with the Brick by Brick Challenge

Join the Global Effort to Automate Building Data Classification

Be part of a transformative challenge to streamline building data management, paving the way for smarter, energy-efficient, and sustainable buildings. This global competition invites participants to develop innovative AI solutions for classifying time-series data from IoT devices using the Brick schema, a step towards redefining building management.

:seedling: ADVANCING ENERGY-EFFICIENT SMART BUILDINGS

Buildings are among the largest consumers of energy worldwide, contributing significantly to global emissions. Managing building systems data is crucial for energy efficiency, but manual processes hinder scalability. The Brick by Brick Challenge addresses this by automating the data classification process, making advanced building management systems accessible and efficient.

Participants will work with time-series data from IoT devices to create models that classify data streams according to the Brick schema. This automation minimises manual effort, enabling scalable solutions that reduce energy consumption and environmental impact.

:scroll: THE TASK

Participants will tackle two core challenges to advance the scalability of smart buildings:

Classify Time-Series Data: Develop models that categorise IoT sensor readings, such as temperature and alarms, into predefined labels based on the Brick schema.

Optimise Data Management: Automate standardisation of diverse building data streams, fostering efficiency and sustainability.

The competition’s multi-label structure enables precise classifications while encouraging innovative solutions for complex building systems.

:bar_chart: DATASET

The dataset includes anonymised time-series data from three Australian buildings. Key features:

Signals: Sensor readings, setpoints, and alarms captured as timestamp-value pairs.

Sampling Rates: Irregular time intervals to simulate real-world scenarios.

Segmentation: Data split into training, leaderboard testing, and competition testing sets for rigorous evaluation.

Labels are based on the Brick ontology v1.2.1, featuring 94 sub-classes, fostering precise and hierarchical data classifications.

:hammer_and_wrench: STARTER KITS AND RESOURCES

Kickstart your participation with the official repository, which includes:

Baseline Models: From naive approaches to tuned Transformers.

Starter Kit: Pre-built tools to make your first submission seamless.

Community Resources: Explore existing classification strategies on Papers with Code.

:chart_with_upwards_trend: EVALUATION CRITERIA

The competition employs a Micro F1 score as the evaluation metric, calculated for each label while accounting for hierarchical relationships. Key rules include:

Zero Labels Masking: Exclude unrelated predictions to ensure fairness.

Specificity Encouraged: Models predicting more precise labels are rewarded, not penalised.

Leaderboard Rankings: Final scores are determined by averaging micro F1 scores across all labels.

:date: TIMELINE

Round 1 Launch: 9th December 2024

Round 2 Launch: 9th January 2025

Competition Close: 3rd February 2025

Winners Announced: 7th February 2025

Presentation: At The Web Conference 2025 (TBC)

:trophy: PRIZES

With a total prize pool of $20,000 AUD, the challenge offers:

Cash Prizes:

:1st_place_medal: 1st Place: $5,000 AUD

:2nd_place_medal: 2nd Place: $3,000 AUD

:3rd_place_medal: 3 x $1,000 AUD

:earth_africa: BUILD A GREENER FUTURE

Take the challenge to automate building data management, making smart, energy-efficient buildings a reality. Join us in reshaping global building management systems for a sustainable future.

Sign up today and transform the way we manage buildings—Brick by Brick!