Stefano Mattoccia, of DEIS will held a lecture at MICC entitled “Stereo vision algorithms for dense 3D reconstruction: introduction and recent developments”.

The lecturer Stefano Mattoccia, DEIS, University of Bologna

The stereo vision enables the 3D reconstruction of scenes observed by two or more cameras. In this seminar, considering the case of dense 3D reconstructions, the main problems of stereo vision will be introduced and recent developments in this area will be examined with particular reference to algorithms that lend themselves to being mapped on devices with parallel processing capabilities (eg FPGA, GPU).

Marco Pedersoli will present a method that can dramatically accelerate object detection with part based models. The method is based on the observation that the cost of detection is likely to be dominated by the cost of matching each part to the image, and not by the cost of computing the optimal configuration of the parts as commonly assumed. Therefore accelerating detection requires minimizing the number of part-to-image comparisons.

A method for the fast inference of multi-resolution part based models. (a) example detections; (b) scores obtained by matching the lowest resolution part (root filter) at all image locations; (c) scores obtained by matching the intermediate resolution parts, only at location selected based on the response of the root part; (d) scores obtained by matching the high resolution parts, only at locations selected based on the intermediate resolution scores

To this end we propose a multiple-resolutions hierarchical part based model and a corresponding coarse-to-fine inference procedure that recursively eliminates from the search space unpromising part placements. The method yields a ten-fold speedup over the standard dynamic programming approach and is complementary to the cascade-of-parts approach. Compared to the latter, our
method does not have parameters to be determined empirically, which simplifies its use during the training of the model. Most importantly, the two techniques can be combined to obtain a very significant speedup, of two orders of magnitude in some cases.

We evaluate our method extensively on the PASCAL VOC and INRIA datasets, demonstrating a very high increase in the detection speed with little degradation of the accuracy.

Prof. Angelo Marcelli will present a principled approach to automatic cursive handwriting reading. The approach is based upon handwriting generation models, and according to them assumes that handwriting is a learned complex motoric task which is accomplished by sequencing simpler movement called stroke.

Automatic reading of cursive handwriting

As learning proceeds in human, so does fluency, which results in producing similar sequence of strokes in correspondence of sequence of letters. Such invariants represents therefore the basic drawing units to which an interpretation can be associated. Reading is then achieved by detecting the invariants used to produce the word to be recognized, associating to them their interpretations, and eventually concatenating the interpretations along the ink of the word, without explicit segmentation in characters, as in case of analytical approaches, or using dictionary of possible words, as in case of holistic methods.

He will conclude with a list of key issues to be addressed while pursuing the proposed approach and the steps we have undertaken along this path.

Understanding driver behavior and ethnography surrounding the task of driving are essential in the development of human-centric driver assistance systems.

Laboratory for Intelligent and Safe Automobiles University of California at San Diego

Novel instrumented vehicles are used for conducting experiments, where the rich contextual information about vehicle dynamics, surround and driver state are captured for careful, detailed ethnographic studies, as well as realistic data for developing algorithms to analyze multi sensory signals for active safety.

In this presentation, Prof. Mohan M. Trivedi will provide a systems- oriented framework for developing multimodal sensing, inferencing algorithms and human-vehicle interfaces for safer automobiles.

He will consider three main components of the system, driver, vehicle, and vehicle surround. He will discuss various issues and ideas for developing models for these main components as well as activities associated with the complex task of safe driving.

The presentation will include discussion of novel sensory systems and learning algorithms for capturing not only the dynamic surround information of the vehicle but also the state, intent and activity patterns of drivers.

He will also introduce a new type of visual display called “dynamic active display”. These displays present visual information to the driver where driving view and safety-critical visual icons are presented to the driver in a manner that minimizes deviation of her gaze direction without adding to unnecessary visual clutter.

These contributions support the practical promise of the “human-centric active safety” (HCAS) systems in enhancing the safety, comfort, and convenience.

For videos and publication list visit LISA

In this talk, Marco Gori introduce the notion of developmental agents, that are based on the theory of “learning from constraints” (see e.g. http://videolectures.net/marco_gori/).

Perceptual and logic constraints

It is claimed that in most interesting tasks, learning from constraints naturally leads to “deep architectures”, that emerge when following the developmental principle of focusing attention on “easy constraints”, at each stage. Interestingly, this suggests that stage-based learning, as discussed in developmental psychology, might not be primarily the outcome of biology, but it could be instead the consequence of optimization principles and complexity issues that hold regardless of the “body”.

In the second part of the talk, he gives insights on the adoption of the proposed framework in computer vision. The proposed functional approach leads naturally to develop different notions of features, the lower level of which are somehow related to classical SIFT. It is pointed out that the adoption of information-theoretic principles are at the basis of the feature generation either at low or high level of the vision computer hierarchy. The functions that are developed are inherently independent of roto-translations and do acquire scale invariance by the minimization of an appropriate entropy-based measure, that is also at the basis of the focus of attention.

Finally, he gives an overview of different constraints emerging at different layers of the hierarchy, and claim that the overall system is expected to work in any visual environment by acting continuously, with no separation between learning and scene interpretation.