Electrical Circuits, Signal and Image Processing Laboratory
Laboratory Details
Laboratory Title: Electrical Circuits, Signal and Image Processing Laboratory
Faculty: Faculty of Engineering
School: Department of Electrical and Computer Engineering
Email: ipratika@ee.duth.gr
Telephone: +30 25410 79586
| Official Gazette of Establishment: |
|---|
Introduction - Presentation
Field of Research
Signal and Image Processing, Computer Vision and Machine learning, AI
Laboratory Members
| Name | Position/Capacity | Telephone | |
|---|---|---|---|
| Ioannis Pratikakis | Professor | ipratika@ee.duth.gr | +30 25410 79586 |
| Nikolaos Mitianoudis | Professor | nmitiano@ee.duth.gr | +30 25410 79572 |
| Ilias Theodorakopoulos | Assistant Professor | iltheodo@ee.duth.gr | +30 25410 79571 |
Featured Research Programs
| Programme Title | Duration | Funding |
|---|---|---|
| CameSense – Advancing precision agriculture through AI-driven monitoring of Camelina sativa crops | 1/4/25 - 31/06/26 | Clusters of Research Excellence (CREs) (ΤΑ 5180519) |
| HAV – Hellenic Autonomous Vehicle | 28/07/2020 - 27/07/2023 | RESEARCH-CREATE-INNOVATE (MIS 5069164) |
| μDoc.tS – Historical Handwritten Document Transcription Platform | 29/06/2018 - 29/06/2021 | RESEARCH-CREATE-INNOVATE (MIS 5030443) |
Featured Publications
| Chronological Order | APP | doi |
|---|---|---|
| 2025 | 10.1016/j.eswa.2025.129158 | |
| 2025 | 10.1109/TIV.2025.3579878 | |
| 2024 | 10.1109/TPAMI.2023.3330944 |
Infrastructure and Equipment
Patents
| Patent Title | Patent Number | Publication Date | Inventor/Company | Summary | URL |
|---|---|---|---|---|---|
| Parsimonious inference on convolutional neural networks | US 12,591,777 B2 | 2026-03-31 00:00:00 | I. Theodorakopoulos et al. / Irida Labs | The patent describes a method for making convolutional neural networks more computationally and energy efficient. A Learning Kernel Activation Module dynamically activates only the kernels or neurons needed for each input during inference, reducing processing time, memory use, and power consumption while maintaining recognition accuracy. This is particularly useful for embedded, IoT, battery-powered, and resource-constrained devices. | https://patents.google.com/patent/US12591777B2/en |
| System and a method to achieve time-aware approximated inference | US 11,526,753 B2 | 2022-12-13 00:00:00 | N. Fragoulis, I. Theodorakopoulos / Irida Labs | The patent describes a neural-network architecture that dynamically adjusts inference complexity according to the available processing time, computational resources, power budget, or required accuracy. Learning Kernel-Activation Modules selectively activate or deactivate convolutional kernels, allowing the system to trade accuracy for faster, lower-power execution, particularly on embedded and resource-constrained devices. | https://patents.google.com/patent/US11526753B2/en |
| System and a method for camera motion analysis and understanding from a video sequence | US 9,508,026 B2 | 2019-01-22 00:00:00 | I. Theodorakopoulos, N. Fragoulis / Irida Labs | The patent describes a system that analyzes consecutive video frames to estimate camera motion and classify the activity or environment of the camera-carrying device. It can recognize motion patterns such as walking, running, shaking, or driving and may also identify the type or identity of the person or object carrying the camera. | https://patents.google.com/patent/US9508026B2/en |
Spin Offs
There are no registered SpinOffs
