Dr Sankaran Iyer

I

I obtained my PhD in 2023 from UNSW, where my research focused on "Vertebral Compression Fracture Detection with a Novel 3D Localization Algorithm." I combined deep reinforcement learning with imitation learning to achieve this. Initially, I used a fully supervised learning approach to predict vertebral compression fractures within localized regions. Later, I explored weakly supervised multiple instance learning and adapted my localization algorithm to work in a semi-supervised setting.

I earned my Master’s degree in Computer Science from UNSW in 1994, with a focus on detecting Latin characters using artificial neural networks.

With over 30 years of industry experience, I have contributed to complex projects related to real-time embedded systems, intelligent networks, and operation support systems. I voluntarily retired in 2016 as a senior project manager at Nokia/Alcatel Lucent.

I have also collaborated with the Biological Earth and Environmental Sciences (BEES) group at UNSW. Together, we developed house dust mite and pest detection systems. Additionally, I contributed to an Android-based app for wildlife species detection as part of the Bushfire Recovery program.

Currently, I work as a senior research associate in collaboration with the Black Dog Institute, focusing on people behavior analysis for suicide detection and prevention. This involves pedestrian detection and tracking in various settings, including GAP parks, railway stations, bridges, and shopping centers.

My areas of interest include object detection and tracking in surveillance areas. Some of my projects include:

• Anomaly detection and tracking in public surveillance systems: Focusing on abandoned object and loitering detection in restricted areas of airports.
• Smart and Secure Parking Lot Management Using Deep Learning: Utilizing license plate recognition and anomaly detection algorithms to identify suspicious activities, such as unauthorized access or prolonged parking.
• Optimizing Inventory Management in Retail Stores Using Computer Vision Technologies: Tracking stock levels, identifying misplaced items, and monitoring shelf conditions in real-time.
• Detection and Tracking of Unauthorized Drones in Restricted Airspace Using Deep Learning: Enhancing airspace security and providing a scalable solution for real-time drone monitoring.
• Developing a system to detect and track unauthorized individuals in restricted zones: Such as government buildings or military bases, to enhance security measures and ensure the safety and integrity of critical infrastructure and personnel.