The research paper titled "Efficient and Concise Explanations for Object Detection with Gaussian-Class Activation Mapping Explainer", led by Nguyễn Quốc Khánh, has been officially published at the 37th Canadian Conference on Artificial Intelligence (CANAI 2024), taking place from May 27–31, 2024, in Guelph, Ontario, Canada.

This milestone marks QAI’s presence in one of the world’s leading AI hubs, demonstrating the deep research capabilities of QAI in AI and Data Science. It also serves as a springboard for collaboration with universities and research institutes, advancing QAI.NSA’s long-term objective of expanding into the North American market.

For Nguyễn Quốc Khánh, a young researcher from the QAI R&D team, this publication is a significant achievement—it is his first first-authored paper accepted at a prestigious international conference. Reflecting on the journey, Khánh shared: "While working on XAI (Explainable AI) algorithms for the akaCam product, I explored various ways to optimize processing time and discovered an efficient approach. With strong encouragement and support from my research team, I was determined to publish this paper, contributing not only to the product but also to the broader scientific community."

G-CAME: A Breakthrough in Explainable AI The paper introduces G-CAME (Gaussian-Class Activation Mapping Explainer), a novel method that significantly reduces the computational time of XAI algorithms. This optimization enhances akaCam Box’s AI processing speed, enabling near-realtime or realtime object detection—a crucial requirement for warning and alert systems. Additionally, the method helps reduce operational costs for customers, making AI deployment more efficient.

Speaking about this achievement, Hùng Nguyễn Trương Thành, Head of QAI.NSA, stated: "I highly value Khánh’s creativity and quick adaptability in various QAI projects and research initiatives. Despite his young age, his expertise spans multiple AI domains, including Computer Vision, Simulator Technology, and Explainable AI, allowing him to bridge the gaps between different fields. Notably, the G-CAME method, featured in his paper, originates from real-world challenges within QAI’s strategic akaCam product—enhancing interpretability and the overall performance of AI models in object detection."

Representing Khánh at CANAI 2024, Nguyễn Trương Thành Hưng emphasized the significant speed improvement achieved with G-CAME. "After overcoming numerous challenges in refining the method, Khánh and the research team successfully reduced the processing time from 250 seconds to just 0.5 seconds—a 400x improvement in generating model explanations. This achievement is a testament to their hard work and dedication. After a year of refinement, the paper was officially accepted and is now being presented at one of Canada’s largest AI conferences."

As part of QAI’s mission in Canada, Thành Hưng also reflected on the strategic objectives set for QAI’s expansion: "Khánh’s success is part of our broader vision to expand into the Canadian AI market, aligning with my OKRs of establishing strong connections with top universities and research institutes worldwide. Within Q1 of 2024 alone, QAI in Canada has successfully engaged with institutions such as the University of New Brunswick (Canada), University of Waterloo (Canada), and the University of Erlangen-Nuremberg (Germany)."

Beyond CANAI 2024, QAI Canada has also achieved additional research milestones, including two papers presented at IEEE ICCE 2024 (Las Vegas, USA, January 2024) and IJCAI 2024 (Jeju, South Korea, August 2024). "While our initial expansion into Canada presents numerous challenges—including complex projects, time zone differences, and geographical distance—I have full confidence in QAI’s expertise to tackle these obstacles and bring our AI solutions to Canada and North America," Thành Hưng added.

About CANAI 2024 The 37th Canadian AI Conference (CANAI 2024) is an annual gathering of top AI researchers from leading institutes and companies, including MILA, CIFAR, and DeepMind. The event serves as a crucial platform for fostering a stronger AI research community in Canada, showcasing Canada’s leadership in intelligent systems and AI innovation.