Compact and efficient deep neural networks for ubiquitous computer vision

Updated 20.10.2023 | 14:10

Computer vision has progressed in many fronts thanks to the adoption of deep neural networks. For example, problems like automatic image classification, segmentation and captioning, as well as object recognition and human pose estimation have benefited from deep neural networks and related recent advances.

Besides academia, also the industry has expressed considerable interest towards neural network based solutions, which can be seen in the form of company acquisitions and research investments. The large-scale deployment of deep neural networks has however been hindered by computational challenges, which limit their portability to mobile platforms and vehicles.

To this extent, the aim of this project is to develop new methodologies for constructing more efficient deep neural networks. The challenge is addressed from two aspects: first, by optimizing existing deep neural networks, and second, by developing efficient neural network architectures from scratch. In addition, the aim is to advance state-of-the-art in several computer vision tasks, including (1) person detection and pose estimation, (2) automatic image captioning, (3) local imagefeature detection and description, and (4) text detection from images.

Name of the project: Compact and efficient deep neural networks for ubiquitous computer vision
Project leader at the University of Vaasa: Jani Boutellier
Leader of the whole project: Juho Kannala, Aalto University
Time: 9/2017-8/2021
External funding from: The Academy of Finland
University of Vaasa's share of the external funding: 133 453 euros (1.1.2020-31.8.2021)
Share of the budget: 190 647 euros (1.1.2020-31.8.2021)
Research Group/ Research Platform: Digital Economy
Contact person at the University of Vaasa: Jani Boutellier
Research partners: Aalto University, University of Maryland
Project number: 334755 (originally 309903)