Publications

In this paper we define a new paradigm for 8-connection labeling, which employes a general approach to improve neighborhood exploration and minimizes the number of memory accesses. Firstly we exploit and extend the decision table formalism introducing OR-decision tables, in which multiple alternative actions are managed. An automatic procedure to synthesize the optimal decision tree from the decision table is used, providing the most effective conditions evaluation order. Secondly we propose a new scanning technique that moves on a 2x2 pixel grid over the image, which is optimized by the automatically generated decision tree.An extensive comparison with the state of art approaches is proposed, both on synthetic and real datasets. The synthetic dataset is composed of different sizes and densities random images, while the real datasets are an artistic image analysis dataset, a document analysis dataset for text detection and recognition, and finally a standard resolution dataset for picture segmentation tasks. The algorithm provides an impressive speedup over the state of the art algorithms.

Nowadays the need for powerful hardware architectures, which allow for high throughput data analysis and calculus, is fundamental especially for biological applications. We have been focused on utilizing the Graphic Processing Unit (GPU) architectures of NVIDIA for accelerating a lipid bilayer simulation tool for biomembranes. ©2010 IEEE.

Contextual information can be used both to reduce computationsand to increase accuracy and this paper presentshow it can be exploited for people surveillance in terms ofperspective (i.e. weak scene calibration) and appearance ofthe objects of interest (i.e. relevance feedback on the trainingof a classifier). These techniques are applied to a pedestriandetector that exploits covariance descriptors througha LogitBoost classifier on Riemannian manifolds. The approachhas been tested on a construction working site wherecomplexity and dynamics are very high, making human detectiona real challenge. The experimental results demonstratethe improvements achieved by the proposed approach.

Pattern recognition is becoming a key application in bodyarea networks. This paper presents a framework promoting unsupervised training for multi-modal, multi-sensor classification systems. Specifically, it enables sensors provided with patter-recognition capabilities to autonomously supervise the learning process of other sensors. The approach is discussed using a case study combining a smart camera and a body-worn accelerometer. The body-worn accelerometer sensor is trained to recognize four user activities pairing accelerometer data with labels coming from the camera. Experimental results illustrate the applicability of the approach in different conditions.

Many works address the problem of object detection by means of machine learning with boosted classifiers. They exploit sliding window search, spanning the whole image: the patches, at all possible positions and sizes, are sent to the classifier. Several methods have been proposed to speed up the search (adding complementary features or using specialized hardware). In this paper we propose a statisticalbased search approach for object detection which uses a Monte Carlo sampling approach for estimating the likelihood density function with Gaussian kernels. The estimation relies on a multi-stage strategy where the proposal distribution is progressively refined by taking into account the feedback of the classifier (i.e. its response). For videos, this approach is plugged in a Bayesian-recursive framework which exploits the temporal coherency of the pedestrians. Several tests on both still images and videos on common datasets are provided in order to demonstrate therelevant speedup and the increased localization accuracy with respect to sliding window strategy using a pedestrian classifier based on covariance descriptors and a cascade of Logitboost classifiers.

This paper proposes a comparison of the two most advanced algorithms for connected components labeling, highlighting how they perform on a soft core SoC architecture based on FPGA. In particular we test our block based connected components labeling algorithm, optimized with decision tables and decision trees. The embedded system is composed of the CMOS image sensor, FPGA, DDR SDRAM, USB controller and SPI Flash. Results highlight the importance of caching and instructions and data cache sizes for high performance image processing tasks.

On-line action recognition from a continuous stream of actionsis still an open problem with fewer solutions proposedcompared to time-segmented action recognition. The mostchallenging task is to classify the current action while findingits time boundaries at the same time. In this paper wepropose an approach capable of performing on-line actionsegmentation and recognition by means of batteries of HMMtaking into account all the possible time boundaries and actionclasses. A suitable Bayesian normalization is appliedto make observation sequences of different length comparableand computational optimizations are introduce to achievereal-time performances. Results on a well known actiondataset prove the efficacy of the proposed method

Proceedings of the XIth International Conference on Artificial Intelligence