Research

3D digital preservation

The IMAGO Research Group has been collaborating with UNESCO and IPHAN in a project that aims at preserving Brazilian baroque sculptures made by artist Antônio Francisco Lisboa, known as O. Aleijadinho, who is considered one of the most important American Baroque artists. It was developed a complete 3D digital preservation pipeline composed by 3D reconstruction and 3D visualization.

*The 3D Virtual Museum is available here.


Assistive Technologies

Our research envelops assistive software and hardware development targeting physically and mentally challenged people, affected with conditions such as autism spectrum disorders (ASDs), visual impairment, and  others. Among the tools we develop there is a facial expression learning software, assisting ASD children to learn how to interpret and reproduce facial expressions with the use of fun 3D animated characters. Another application, also aimed at ASD children, tracks the user eyes movements to assess what visual information interests him/her the most, as it is known that people with ASD present a different form of observing the surroundings. For the visual impaired we developed an haptic hardware, called haptic mouse, paired with a software, that recreates a 3D virtual object to be experienced through the user’s touch, enabling the recreation of those objects inside the mind.


Face analysis

Another one of our research areas aims at face analysis  in controlled and in-the-wild environments. The potential of the nose region is explored for computing the face pose in difficult scenarios. This estimative can be directly applied to increase face tracking performance and providing a useful initial guess of the position of facial landmarks. Also, in the context of face analysis, facial expressions play an important role in human communication and day-to-day life. Because of this, we are working on computer vision solutions to understand these expressions through Action Units (the building blocks of facial expressions) detection and intensity estimation.


Biometrics

Our research involves the development of new biometric  identification methods using innovative features present in high-resolution 3D fingerprints scanned through an Optical Coherence Tomography  (OCT) sensor.  Anatomic features such as the pores or minutiae can be used and geometric features such as the curvatures (Gaussian and Mean) can be computed. Our 3D OCT database contains 3D fingerprints from the epidermis and the dermis and was extracted from volunteers of diverse  ethnicities, including cases of alterations as scars and calluses. Extraction of different regions of interest (ROI) such as regions round the pores (pore clouds) or around the minutiae (minutiae clouds) permit several matching scores as the Hausdorff between clouds or Euclidean distances between texture  descriptors (local gradient patterns). Our methods are robust to fingerprint alterations by exploiting the internal 3D fingerprints from the dermis-epidermis interface.

3D Fingerprint

3D fingerprint