1. Developed mobile phone friendly machine learning model to predict the quality of Segmentation in AR applications and demonstrated its effectiveness in terms of accuracy, latency and memory usage.

2. Demonstrated the utility of simple computer vision techniques in determining the spatial drift experienced in mobile AR applications. Utilized key insights from the study to develop a method to reduce the spatial drift in AR applications.

3. Developed adaptive method that dynamically shifts between Visual Odometry and Simultaneous Localization and Mapping (SLAM) based on the visibility of the virtual object, to enable more memory and latency efficient MAR application, while maintaining the quality of the AR experience.

4. Currently working on developing a method to perform SLAM in dynamic environments using Transformers.

5. Developed a method that uses mobile phone friendly machine learning model to dynamically shifts between template matching and segmentation to enable real-time dynamic SLAM in MAR applications.

Associated Publications

1. S. M. K. Swamy and Q. Han, “Quality evaluation of image segmentation in mobile augmented reality,” in EAI International Conference on Mobile and Ubiquitous Systems: Computing, Networking and Services (Mobiquitous), 2023

2. S. M. K. Swamy and Q. Han, “Online mitigation of spatial drift of virtual objects in mobile augmented reality,” Submitted to Workshop in SigComm, 2024.

3. S. M. K. Swamy and Q. Han, “Real-time dynamic slam using rgb cameras for mobile augmented reality,” Submitted to Workshop in SigComm, 2024.

4. S. M. K. Swamy and Q. Han, “Appear: Adaptive pose estimation for mobile augmented reality,” Submitted to ISMAR, 2024.