Machine Vision and Perceptual Intelligence Laboratory (PILAB) was founded in 2008 under Işık University Computer Science and Engineering Department. Our research focuses on theoretical and applied image processing and computational intelligence.

PILAB team is composed of 4 faculty, one PhD student and 3 MS students. We are looking for MS/PhD students who will study the field of Augmented Reality.

We are also open to all Işık University undergraduate students who would like to participate in our projects. See our flyer for ongoing projects.

The themes of our current projects are:

• Facial Expression Synthesis and Analysis (funded 2009-2012)
• Autonomous Vehicles
• Industrial Quality Control applications
• Augmented Reality

Facial Expression Analysis and Synthesis (Funded by Tübitak 2009-2012): Human face shelters most of the sense organs and important expression mechanisms. Face serves as an interface for one of the principal modes of interpersonal communication. Our goal is to track the emotion of a face in a video and determine the muscle activations. The muscle activations that compose a facial expression are solved under the constraints of an anatomical face model. Using these muscle activations we will be able to automatically detect subtle facial expressions. These expressions give us clues about the emotional state of the subject, such as happiness, surprise, anger, sadness, nervousness and deception.

Neural Network Ensembles: Artificial neural networks is one of the oldest fields of Artificial Intelligence. Inspired by the human brain, an artificial neural network is composed of neurons, each firing electrical signals over axons to other connected neurons. A neuron will be active and fire only if it is stimulated to sufficient level by other neurons or sensory inputs. Artificial neural networks have been used successfully in many real life application areas such as game playing, vehicle control and financial predictions. In this research project, we are testing if previously learned concepts would enhance learning of brand new concepts. We first train individual neural networks on simpler problems and create an ensemble of these networks to test its performance in dealing with more complex problems.

Real-Time Defect Detection of Pantographs in Rail Transport: Pantograph is the device that conducts electricity from the catenary wires to the train. The catenary wires that hover above the train may suffer from metal fatigue and yield or produce bulges. Its impact on a train that travels at 120 km/h can be catastrophic.

These impacts could break the pantograph, even worse the catenary wire, which brings all trains on the same route to standstill. In this project we installed a high-resolution camera on an inter-city train to collect data. We are developing an image processing application to automatically detect electrical arcs, impacts and defects on the pantograph. This information will be instantly transmitted to the central control office over the GSM-R (Global System for Mobile Communications – Railway) network. Early detection of light impacts and small defects will prevent catastrophic failures.

Contact us for more information about the PILAB.