Roberto Lam

Cortical 3D Face and Object Recognition Using 2D Projections

Empirical studies concerning face recognition suggest that faces may be stored in memory by a few canonical representations. In cortical area V1 exist double-opponent colour blobs, also simple, complex and end-stopped cells which provide input for a multiscale line/edge representation, keypoints for dynamic feature routing, and saliency maps for Focus-of-Attention. All these combined allow faces to be segregated. Events of different facial views are stored in memory and combined to identify the view and recognise a face, including its expression. In this paper, the authors show that with five 2D views and their cortical representations it is possible to determine the left-right and frontal-lateral-profile views, achieving a view-invariant recognition rate of 91%. The authors also show that the same principle with eight views can be applied to 3D object recognition when they are mainly rotated about the vertical axis.

Cortical multiscale line-edge disparity model

Most biological approaches to disparity extraction rely on the disparity energy model (DEM). In this paper we present an alternative approach which can complement the DEM model. This approach is based on the multiscale coding of lines and edges, because surface structures are composed of lines and edges and contours of objects often cause edges against their background. We show that the line/edge approach can be used to create a 3D wireframe representation of a scene and the objects therein. It can also significantly improve the accuracy of the DEM model, such that our biological models can compete with some state-of-the-art algorithms from computer vision.

Retrieval of 3D polygonal objects based on multiresolution signatures

In this paper we present a method for retrieving 3D polygonal objects by using two sets of multiresolution signatures. Both sets are based on the progressive elimination of objects details by iterative processing of the 3D meshes. The first set, with five parameters, is based on mesh smoothing. This mainly affects an objects surface. The second set, with three parameters, is based on difference volumes after successive mesh erosions and dilations. Characteristic feature vectors are constructed by combining the features at three mesh resolutions of each object. In addition to being invariant to mesh resolution, the feature vectors are invariant to translation, rotation and size of the objects. The method was tested on a set of 40 complex objects with mesh resolutions different from those used in constructing the feature vectors. By using all eight features, the average ranking rate obtained was 1.075: 37 objects were ranked first and only 3 objects were ranked second. Additional tests were carried out to determine the significance of individual features and all combinations. The same ranking rate of 1.075 can be obtained by using some combinations of only three features.

GyGSLA: A Portable Glove System for Learning Sign Language Alphabet

The communication between people with normal hearing with those having hearing or speech impairment is difficult. Learning a new alphabet is not always easy, especially when it is a sign language alphabet, which requires both hand skills and practice. This paper presents the GyGSLA system, standing as a completely portable setup created to help inexperienced people in the process of learning a new sign language alphabet. To achieve it, a computer/mobile game-interface and an hardware device, a wearable glove, were developed. When interacting with the computer or mobile device, using the wearable glove, the user is asked to represent alphabet letters and digits, by replicating the hand and fingers positions shown in a screen. The glove then sends the hand and fingers positions to the computer/mobile device using a wireless interface, which interprets the letter or digit that is being done by the user, and gives it a corresponding score. The system was tested with three completely inexperience sign language subjects, achieving a 76 % average recognition ratio for the Portuguese sign language alphabet.

Using mathematical morphology for similarity search of 3D objects

In this paper we use the erosion and dilation operators for characterizing 3D polygonal objects. The goal is to perform a similarity search in a set of distinct objects. The method applies successive dilations and erosions of the meshes in order to compute the difference volume as a function of the size of the structuring element. Because of appropriate pre-processing, the resulting function is invariant to translation, rotation and mesh resolution. On a set of 32 complex objects with different mesh resolutions, the method achieved an average ranking rate of 1.47, with 23 objects ranked first and 6 objects ranked second.

Mobile Augmented Reality Framework - MIRAR

The increasing immersion of technology on our daily lives demands for additional investments in various areas, including, as in the present case, the enhancement of museums’ experiences. One of the technologies that improves our relationship with everything that surrounds us is Augmented Reality. This paper presents the architecture of MIRAR, a Mobile Image Recognition based Augmented Reality framework. The MIRAR framework allows the development of a system that uses mobile devices to interact with the museum’s environment, by: (a) recognizing and tracking on-the-fly, on the client side (mobile), museum’s objects, (b) detecting and recognizing where the walls and respective boundaries are localized, as well as (c) do person detection and segmentation. These objects, wall and person segmentation will allow the projection of different contents (text, images, videos, clothes, etc.). Promising results are presented in these topics, nevertheless, some of them are still in a development stage.

Expression-invariant face recognition using a biological disparity energy model

Abstract Biologically–compatible methods are not commonly used for face recognition. Complex computational approaches are preferred and dominate the state of the art. However, we know that the human brain is very efficient at processing faces, without explicitly depending on advanced mathematics. In this paper we focus on evaluating the performance of an expression-invariant face recognition system, which is based on the most widely-accepted biological model of stereo vision: the Disparity Energy Model (DEM), which has been shown to deliver precise but inaccurate results. We show that the DEM can provide 3D disparity maps which are suitable for both identity recognition and verification, even coping with a wide range of facial expressions. We test disparity information, both alone and in combination with luminance data, achieving state-of-the-art results. We also compare DEM results with those obtained by precise and accurate laser range maps, concluding that the differences in performance are very small.