LESLearn: Label, Energy and Sample Efficient Feature Representation and Learning in Computer Vision
Description of the granted funding
We are moving toward a future where everything will be connected and intelligent, simplifying and enriching our daily lives. AI techniques, featured by Deep Neural Networks (DNNs), are driving this revolution. The aim of this project is to create the technology needed for learning visual feature representations in an energy efficient, human-like manner, overcoming the limitations of DNNs in being energy and data hungry. The first goal is to develop energy efficient DNNs without sacrificing application accuracy or increasing hardware cost. The second goal is to develop effective few shot learning methods which enable learning by using as little labeled training data as people need and can rapidly generalize to a range of different tasks, narrowing the gap between AI and real human intelligence. The outcomes of the project will significantly benefit AI powered edge devices to support many computer vision tasks that are set to transform industries and change our lives for the better.
Show moreStarting year
2020
End year
2024
Granted funding
Other information
Funding decision number
331883
Fields of science
Computer and information sciences
Research fields
Tietojenkäsittelytieteet
Identified topics
artificial intelligence, machine learning