Gary Leung

I am a MSc. student at University of Toronto advised by Prof. Sanja Fidler.

Previously, I graduated from University of Toronto in 2020 majoring in Engineering Science (Robotics) and have held internships at Samsung Toronto AI Centre and Google

Outside of work, I enjoy spending my time reading, learning new languages, or gaming.

Email / CV / Google Scholar / Twitter / Github

Research

	Improving Semantic Segmentation in Transformers using Hierarchical Inter-Level Attention Gary Leung, Jun Gao, Xiaohui Zeng, Sanja Fidler arXiv / Project Website We present Hierarchical Inter-Level Attention (HILA), an attention-based method that captures Bottom-Up and Top-Down Updates between features of different levels in hierarchical vision transformer architectures.
	Studying BatchNorm Learning Rate Decay on Meta-Learning Inner-Loop Adaptation Alexander Wang, Gary Leung, Sasha Doubov* NeurIPS Workshop on Meta-Learning, 2021 Open Review We study the effects of BatchNorm's implicit learning rate decay on the inner loop adaptation for MAML. We look to address potential causes of the feature reuse/rapid adaptation in meta-learning.
	VASTA: A Vision and Language-assisted Smartphone Task Automation System Alborz Rezazadeh Sereshkeh, Gary Leung, Krish Perumal, Caleb Phillips, Minfan Zhang, Afsaneh Fazly, Iqbal Mohomed ACM IUI, 2020 arXiv / demo We build VASTA, a vision and language-assisted Programming By Demonstration (PBD) system for smartphone task automation.

Website credit to: ✩