Tengda Han

Tengda Han is a Research Scientist at Google DeepMind, specializing in neural networks for video understanding. He earned his PhD from the University of Oxford under Andrew Zisserman and a BEng in Mechanical & Material Engineering from the Australian National University. Prior to his PhD, he studied business administration and law at Renmin University of China for one year.

• PhD in Computer Science, University of Oxford (2022)
• BEng in Mechanical & Material Engineering, Australian National University (2016)
• One year of business administration and law at Renmin University of China

His research focuses on video representation learning, self-supervised methods, multimodal learning, and efficient prompting techniques. He has pioneered training-free frameworks for audio description, zero-shot object counting, and hyper-realistic movie generation.

Recent publications demonstrate trends in temporal alignment, discriminative prompting, 5D video modeling, and open-world counting. His work often bridges computer vision with natural language processing and speech recognition.

Scientific Awards:

• Best Paper Award at ACCV2024
• Best Poster Award at BMVC2023
• BMVA Sullivan Doctoral Thesis Prize Runner-up

He has co-organized workshops at ICCV25, NeurIPS23, and BMVA Symposium on Vision & Language. Collaborators include Andrew Zisserman, Dima Damen, and Flamingo project members at DeepMind.