By Barry Norton, VP of Human-Centred AI at Milestone Systems.
Video technology has advanced significantly in recent decades, particularly with improvements in video analytics and AI. As the global AI market is projected to soar to USD 1.3 trillion by 2030, according to MarketsandMarkets, one of the major challenges hindering this growth is access to large, high-quality datasets for AI training. Synthetic data is emerging as a crucial solution.
The Role of Synthetic Data in AI Development
Groundbreaking work in AI by figures like 2018 Turing Award winners Yoshua Bengio, Geoffrey Hinton, and Yann LeCun, alongside Fei-Fei Li’s development of ImageNet, has laid the foundation for modern AI applications, especially in computer vision. This is particularly relevant for video technology, which plays a vital role in enhancing the safety and efficiency of Australian cities, transport systems, retail environments, and public spaces.
With AI, organisations can extract valuable insights to make data-driven decisions—whether it’s planning road infrastructure, optimising retail layouts, or improving maintenance scheduling. The fusion of AI and video analytics is driving smarter, safer environments across Australia.
The Challenge of Training AI Models
For AI models to perform accurately, they require extensive datasets that are diverse, representative, and ethically sourced. Obtaining such data is often challenging, particularly for AI applications involving video, as cameras can capture personal and sensitive information. Privacy concerns, data security, and regulatory restrictions can limit the availability of real-world datasets, slowing AI development.
This is where synthetic data is proving to be a game-changer.
How Synthetic Data Addresses These Challenges
Synthetic data consists of artificially generated or augmented datasets that mimic real-world conditions. It enables AI developers to train models on vast, varied data while mitigating ethical, legal, and privacy concerns. By preserving critical real-world characteristics without exposing individuals to privacy risks, synthetic data accelerates AI model development.
Additionally, synthetic data can help reduce bias in AI models. Traditional datasets often reflect historical biases, potentially leading to skewed AI decision-making. By carefully designing synthetic data, developers can create more balanced and fair datasets.
Moreover, synthetic data is scalable and cost-effective, allowing AI developers to generate high-quality datasets rapidly and affordably. This is particularly beneficial for AI applications requiring highly specific training data that is difficult to obtain in large quantities.
Australian Case Study: Enhancing Water Safety
The potential of synthetic data in life-saving AI applications is evident in a research initiative in Denmark aimed at preventing drownings in harbours. A similar approach could be adopted in Australia, where drowning prevention remains a significant public safety issue.
Between 2001 and 2015, 1,647 lives were lost in Danish waters, with a quarter of those occurring in harbours. In Aalborg Harbour, researchers trained AI-equipped video cameras to detect when a person fell into the water using a thermal dataset. While human volunteers participated in controlled falls, it was impractical and unsafe to simulate real accidental falls, particularly for wheelchair users, cyclists, and skateboarders.
By incorporating synthetic data, researchers expanded their training dataset, enabling AI models to recognise diverse falling scenarios while ensuring safety. The AI system developed through this project showed promise in alerting rescue teams in real-time, reducing response times, and potentially saving lives. This approach could be highly relevant for Australian coastal cities and harbours, where AI-driven video analytics could enhance water safety.
Broader Applications for Synthetic Data in Australia
Beyond water safety, synthetic data has widespread applications across various industries in Australia:
- Manufacturing: AI-driven quality control and predictive maintenance can prevent costly downtime, but obtaining real-world data from production lines can be risky due to proprietary information. Synthetic data can bridge this gap.
- Healthcare: Protecting patient privacy is paramount, making it difficult to collect real-world training data for AI applications such as fall detection in hospitals and aged care facilities. Synthetic data enables AI to learn from realistic scenarios without compromising privacy.
- Retail and Smart Cities: AI models can optimise store layouts and urban infrastructure, but real-world data collection often raises privacy concerns. Synthetic data allows for AI training without breaching data protection laws.
A Growing Opportunity for AI in Australia
As AI adoption increases across Australian industries, the use of synthetic data is set to grow in parallel. Providing a scalable, ethical, and privacy-compliant data source, synthetic data is becoming a critical enabler of AI development. Organisations leveraging video analytics should consider synthetic data as a powerful tool to enhance AI accuracy, fairness, and overall effectiveness.
By embracing synthetic data, Australia can continue to innovate in AI-driven video analytics while upholding high standards of privacy, security, and ethical AI development.
