Your visit here means that you are definitely curious...
In the following, you can find demos and short descriptions of the work carried out in the VIPS lab in the context of projects or theses of Master and PhD students.
Hope you can find them interesting!
MY 3D FACE MODELING
This model was done during my visit in the IRIS lab of
Thanks to P. Mordohai (IRIS,
3D face manipulation (VRML plug-in required)
VIPS PEOPLE FACE MORPHING
This is a video containing the morphing of the VIPS people faces.
Credit: M. Farenzena
This is a Master thesis investigating the face recognition problem. It was tackled by modelling faces using the multi-level B-spline interpolation technique and using the resulting coefficients to feed a set of Support Vector Machines organized in a tree fashion.
Credit: G. Iacono
REGISTRATION OF HISTORIC AND CURRENT AERIAL IMAGES
This work proposes a semi-automatic application for the alignment of historical aerial images and current aerial images in order to improve the accuracy and speed-up alignment process.The task is challenging because a lot of features in the historical images are changed or missing (and vice versa), and acquisition parameters are not controlled (scale, orientation, etc.).
Credit: A. Etrari
Link (PDF, 1734 KB)
(PDF, 1734 KB)
This work proposes a technique for tracking complex objects (both polyhedral and smooth boundaries) in a monocular sequence. The aim is to use this model tracking method in an Augmented Reality context to compute the pose of a real object to be able to register it with a synthetic one. This technique is robust to occlusions, since the whole object contour is used, not just few control points. The proposed method is effective yet simple: no image feature extraction is necessary and no complex temporal evolution is used.
Credit: A. Valinetti
VIEW SYNTHESIS FROM UNCALIBRATED SPARSE IMAGES
This work proposes an image-based system for novel view synthesis from multiple model views. The method works by segmenting images of a static scene in background and foreground, on the basis of motion parallax. From this segmentation, we are able to recover the relative affine structure,and, finally, novel views are synthesized using an original method based on step-wise replication of the epipolar geometry acquired from few model or “seed” views. The method is uncalibrated, for it does not need the rigid displacements in the Euclidean frame (which is unknown), and it is automatic, for it does not require the user to manually specify viewing parameters.
Credits: S. Caldrer, S. Ceglie
VIEW SYNTHESIS FROM A SINGLE IMAGE
This work proposes
a method for generating synthetic views of a soccer ground starting from a
single uncalibrated image. The relative affine structure of the players is
computed by exploiting the knowledge of the soccer ground geometry and the fact
that the players are in vertical positions. Then, novel views are generated
using the "plane+parallax" representation to reproject points.
Credit: N. Mattern
ELEMENTARY SHAPE RECOGNITION USING EIGENSPACES
This work aims at investigating a method for the recognition of elementary shapes in a single image. It is assumed that a large set of man-made objects can be quite faithfully approximated by a small set of standard primitives composed of single geometric shapes such as rectangles or circles. The objective of this work is thus to detect generic objects when their surfaces resemble one model primitive.
Credits: A. Maschi
MOSAICING AND LAYERED REPRESENTATIONS
This work proposes an application which takes a video shot as input and produces an object-based representation composed by a background (still) mosaic and moving regions. First, an high quality seamless mosaic of background is produced. The optimal alignment is achieved with a graph-based technique that exploits the information coming from every overlapping frame pair. Moving regions are segmented using a model of the background derived from robust statistics. Tracking is based on blob matching on a graph representation of the sequence. Specific techniques are introduced to cope with crossing and occlusions. Results on mosaicing and coding are reported and some exampled of video-editing as well.
Credits: R. Marzotto, M. Aprile
3D ENVIRONMENT MODELING FROM UNDERWATER ACOUSTIC IMAGES
This work proposes a technique for the three-dimensional reconstruction of an underwater environment from multiple range views. The final target lies in improving the understanding of a human operator guiding an underwater remotely operated vehicle (ROV) equipped with an acoustic camera, which provides a sequence of 3D images in real time. Since the field of view is narrow, we devise a technique for the reconstruction of relevant information of the image sequence up to build a mosaic of the surrounding scene. Due to the very noisy nature of the data and the low range resolution, smoothing, segmentation, registration, and fusion problems have been tackled. Examples on real images have been presented to show the promising performances of the algorithm.
Credit: L. Ronchetti
Link (PDF, 1054 KB)
(PDF, 1054 KB)
2D SHAPE RECOGNITION USING HIDDEN MARKOV MODELS
In this work, Hidden Markov Models (HMMs) are investigated for the purpose of classifying planar shapes represented by their curvature coefficients. In the training phase, special attention is devoted to the initialization and model selection issues, which make the learning phase particularly effective. The results of tests on different data sets show that the proposed system is able to accurately classify objects that were translated, rotated, occluded, or deformed by shearing, also in the presence of noise.
Credit: M. Bicego
CLASSIFICATION OF EEG SIGNAL BY HIDDEN MARKOV MODELS
A method for the clustering of sequential data using an Hidden Markov Model (HMM)-based technique has been proposed. Its application to electro-encephalographic (EEG) signals to identify or classify mental states has been tested.
Credit: A. Panuccio
Link (PDF, 143 KB)
(PDF, 143 KB)
The problem of classification of the defects occurring in a textile manufacture is addressed, and a new classification scheme is devised in which different features, extracted from the gray level histogram, shape, and co-occurrence matrices, are employed. All of these features are classified using a Support Vector Machines (SVM) based framework. An accurate analysis of different multi-class classification schemes and SVM parameters has been carried out. The system has been tested using two textile databases, showing very promising results.
Credit: I.A. Rossi