Advancing AI-Driven Video Analytics at CUEE MDAP Research Unit
At Chulalongkorn University’s Electrical Engineering Department, the Multimedia Data Analytics and Processing (MDAP) Research Unit is actively engaged in developing computer vision techniques with a specific focus on surveillance applications. The unit is dedicated to advancing deep artificial intelligence and machine learning solutions, making significant contributions to the field of multimedia data analytics.
The primary research areas of the lab include:
- Video Anomaly Detection
- Digital Image and Video Super-resolution Techniques
- Digital Video Coding
- Face Recognition and Emotional Expression
One of MDAP’s most prominent research areas is anomaly detection in video surveillance systems, essential for maintaining safety in complex, dynamic environments.
🔎 Highlighted Research: Enhancing Anomaly Detection in Surveillance with Generative Adversarial Networks
Surveillance systems operating in crowded and dynamic environments often encounter challenges, including high rates of false alarms and missed detections. To address these critical issues, MDAP researchers have innovatively applied Generative Adversarial Networks (GANs). This pioneering approach significantly enhances anomaly detection capabilities within these complex scenarios.
Here’s how MDAP’s GAN-based approach revolutionizes anomaly detection:
- Learning Normal Behavior: GANs learn exclusively from typical scenarios to recognize usual patterns.
- Spotting Deviations: Any anomaly is quickly identified by comparing live scenes to the learned normal patterns.
- Precision Localization: Techniques such as Dense Inverse Search, Optical Flow, and edge wrapping enhance accuracy, pinpointing unusual events with sharp precision and speed.
🚀 Real-World Impact
Tested rigorously on recognized benchmarks such as the UCSD pedestrian datasets, this method has delivered remarkable results:
- High Accuracy: Frame-level detection accuracy of 98.5%.
- Precise Localization: Pixel-level accuracy reached 77.4%.
- Efficient Operation: Real-time performance at 30 frames per second, ideal for practical deployment.
Beyond these benchmarks, anomaly detection technology has significant real-world applications with impactful implications:
Rescue Drone
AI-powered drones quickly identify survivors in marine incidents, like shipwrecks or airplane water landings.
Event Safety Monitoring
Rapidly detects suspicious activities at crowded events, improving safety and security.
Smart Traffic Management
Instantly detects accidents and unusual traffic, speeding up emergency response.
🤝 Partnership for Cutting-Edge AI IP Core Development
The MDAP Research Unit and Design Gateway Co., Ltd. have formed a strategic partnership to deliver cutting-edge AI solutions to the semiconductor market. This collaboration leverages MDAP’s profound expertise in AI analytics and advanced digital IC design with Design Gateway’s robust capabilities in developing IP cores for FPGA and ASIC devices. This synergy is poised to achieve several key objectives:
- Accelerated Innovation: Swiftly transform advanced AI research and design expertise into market-ready IP cores, significantly reducing time-to-market.
- Enhanced Performance: Develop industry-leading, high-performance AI IP cores meticulously optimized for FPGA and ASIC applications, ensuring superior efficiency and capability.
- Global Competitiveness: Establish new industry benchmarks by creating high-performance AI IP cores that drive global competitiveness in the semiconductor landscape.
To elevate your products and services, we encourage you to leverage this strategic collaboration. Discover more about the MDAP Research Unit and explore potential partnerships by visiting their website or the dedicated collaboration portal at MDAP x DG.
Explore more exciting projects from MDAP Lab, including innovative senior projects, by visiting their dedicated Senior Project 2025 page (Thai Language Content).
