Kaiyue Pang

Researcher at SketchX

thatkpang(at)gmail.com

Research:

I love pixels and I love people — in my work, I try to bring them together. By developing computer vision and machine learning techniques as a vehicle to process the abstract, subtle and subjective qualities of the intense visual data around us, my research often falls into three long-term goals.

- Improving human resonance and well being in different visual activities, be it creative, educational, emotional or commercial.

- Helping the world to become a better place for mankind by developing protective, automated and robust visual systems for biodiversity monitoring or earth observation.

- Injecting the moral and interpretable dimension into the visual tools so that they would ultimately appreciate Picasso's cubism portraits over Kardashian's selfies, and explain why.

Methodologically, I believe to how well we can approach these goals depends on how able an intelligent visual system is in giving satisfying answers to the following general questions:

- What is the best form of user input to carry the intended purpose of a system?

- Can the system achieve efficient learning under minimal human supervision, and explain and assess its own accuracy?

- Does the system ask questions of users, when it's uncertain?

- How the system evolves over time by digesting human prior knowledge and instruction?

I support Slow Science.