Publications

Event cameras, neuromorphic devices that naturally respond to brightness changes, have multiple advantages with respect to traditional cameras. However, the difficulty of applying traditional computer vision algorithms on event data limits their usability. Therefore, in this paper we investigate the use of a deep learning-based architecture that combines an initial grayscale frame and a series of event data to estimate the following intensity frames. In particular, a fully-convolutional encoder-decoder network is employed and evaluated for the frame synthesis task on an automotive event-based dataset. Performance obtained with pixel-wise metrics confirms the quality of the images synthesized by the proposed architecture.

When you see a person in a crowd, occluded by other persons, you miss visual information that can be used to recognize, re-identify or simply classify him or her. You can imagine its appearance given your experience, nothing more. Similarly, AI solutions can try to hallucinate missing information with specific deep learning architectures, suitably trained with people with and without occlusions. The goal of this work is to generate a complete image of a person, given an occluded version in input, that should be a) without occlusion b) similar at pixel level to a completely visible people shape c) capable to conserve similar visual attributes (e.g. male/female) of the original one. For the purpose, we propose a new approach by integrating the state-of-the-art of neural network architectures, namely U-nets and GANs, as well as discriminative attribute classification nets, with an architecture specifically designed to de-occlude people shapes. The network is trained to optimize a Loss function which could take into account the aforementioned objectives. As well we propose two datasets for testing our solution: the first one, occluded RAP, created automatically by occluding real shapes of the RAP dataset created by Li et al. (2016) (which collects also attributes of the people aspect); the second is a large synthetic dataset, AiC, generated in computer graphics with data extracted from the GTA video game, that contains 3D data of occluded objects by construction. Results are impressive and outperform any other previous proposal. This result could be an initial step to many further researches to recognize people and their behavior in an open crowded world.

Diplegia is a specific subcategory of the wide spectrum of motion disorders gathered under the name of cerebral palsy. Recent works proposed to use gait analysis for diplegia classification paving the way for automated analysis. A clinically established gait-based classification system divides diplegic patients into 4 main forms, each one associated with a peculiar walking pattern. In this work, we apply two different deep learning techniques, namely, multilayer perceptron and recurrent neural networks, to automatically classify children into the 4 clinical forms. For the analysis, we used a dataset comprising gait data of 174 patients collected by means of an optoelectronic system. The measurements describing walking patterns have been processed to extract 27 angular parameters and then used to train both kinds of neural networks. Classification results are comparable with those provided by experts in 3 out of 4 forms.

Novelty detection is commonly referred to as the discrimination of observations that do not conform to a learned model of regularity. Despite its importance in different application settings, designing a novelty detector is utterly complex due to the unpredictable nature of novelties and its inaccessibility during the training procedure, factors which expose the unsupervised nature of the problem. In our proposal, we design a general framework where we equip a deep autoencoder with a parametric density estimator that learns the probability distribution underlying its latent representations through an autoregressive procedure. We show that a maximum likelihood objective, optimized in conjunction with the reconstruction of normal samples, effectively acts as a regularizer for the task at hand, by minimizing the differential entropy of the distribution spanned by latent vectors. In addition to providing a very general formulation, extensive experiments of our model on publicly available datasets deliver on-par or superior performances if compared to state-of-the-art methods in one-class and video anomaly detection settings. Differently from prior works, our proposal does not make any assumption about the nature of the novelties, making our work readily applicable to diverse contexts.

Despite the popularity that saliency models have gained in the computer vision community, they are most often conceived, exploited and benchmarked without taking heed of a number of problems and subtle issues they bring about. When saliency maps are used as proxies for the likelihood of fixating a location in a viewed scene, one such issue is the temporal dimension of visual attention deployment. Through a simple simulation it is shown how neglecting this dimension leads to results that at best cast shadows on the predictive performance of a model and its assessment via benchmarking procedures.

Predicting the Driver's Focus of Attention: the DR(eye)VE Project Andrea Palazzi, Davide Abati, Simone Calderara, Francesco Solera, Rita Cucchiara (Submitted on 10 May 2017 (v1), last revised 6 Jun 2018 (this version, v3)) In this work we aim to predict the driver's focus of attention. The goal is to estimate what a person would pay attention to while driving, and which part of the scene around the vehicle is more critical for the task. To this end we propose a new computer vision model based on a multi-branch deep architecture that integrates three sources of information: raw video, motion and scene semantics. We also introduce DR(eye)VE, the largest dataset of driving scenes for which eye-tracking annotations are available. This dataset features more than 500,000 registered frames, matching ego-centric views (from glasses worn by drivers) and car-centric views (from roof-mounted camera), further enriched by other sensors measurements. Results highlight that several attention patterns are shared across drivers and can be reproduced to some extent. The indication of which elements in the scene are likely to capture the driver's attention may benefit several applications in the context of human-vehicle interaction and driver attention analysis.

State-of-the-art Computer Vision pipelines show poor performances on artworks and data coming from the artistic domain, thus limiting the applicability of current architectures to the automatic understanding of the cultural heritage. This is mainly due to the difference in texture and low-level feature distribution between artistic and real images, on which state-of-the-art approaches are usually trained. To enhance the applicability of pre-trained architectures on artistic data, we have recently proposed an unpaired domain translation approach which can translate artworks to photo-realistic visualizations. Our approach leverages semantically-aware memory banks of real patches, which are used to drive the generation of the translated image while improving its realism. In this paper, we provide additional analyses and experimental results which demonstrate the effectiveness of our approach. In particular, we evaluate the quality of generated results in the case of the translation of landscapes, portraits and of paintings coming from four different styles using automatic distance metrics. Also, we analyze the response of pre-trained architecture for classification, detection and segmentation both in terms of feature distribution and entropy of prediction, and show that our approach effectively reduces the domain shift of paintings. As an additional contribution, we also provide a qualitative analysis of the reduction of the domain shift for detection, segmentation and image captioning.

We address the deployment of perceptual attention to social interactions as displayed in conversational clips, when relying on multimodal information (audio and video). A probabilistic modelling framework is proposed that goes beyond the classic saliency paradigm while integrating multiple information cues. Attentional allocation is determined not just by stimulus-driven selection but, importantly, by social value as modulating the selection history of relevant multimodal items. Thus, the construction of attentional priority is the result of a sampling procedure conditioned on the potential value dynamics of socially relevant objects emerging moment to moment within the scene. Preliminary experiments on a publicly available dataset are presented.

Gene fusions have a very important role in the study of cancer development. In this regard, predicting the probability of protein fusion transcripts of developing into a cancer is a very challenging and yet not fully explored research problem. To this date, all the available approaches in literature try to explain the oncogenic potential of gene fusions based on protein domain analysis, that is cancer-specific and not easy to adapt to newly developed information. In our work, we choose the raw protein sequences as the input baseline, and propose the use of deep learning, and more specifically Convolutional Neural Networks, to infer the oncogenity probability score of gene fusion transcripts and to group them into a number of categories (e.g., oncogenic/not oncogenic). This is an inherently flexible methodology that, unlike previous approaches, can be re-trained with very less efforts on newly available data (for example, from a different cancer). Based on experimental results on a large dataset of pre-annotated gene fusions, our method is able to predict the oncogenity potential of gene fusion transcripts with accuracy of about 72%, which increases to 86% if we consider the only instances that are classified with a high confidence level.

In a weakly-supervised scenario object detectors need to be trained using image-level annotation alone. Since bounding-box-level ground truth is not available, most of the solutions proposed so far are based on an iterative, Multiple Instance Learning framework in which the current classifier is used to select the highest-confidence boxes in each image, which are treated as pseudo-ground truth in the next training iteration. However, the errors of an immature classifier can make the process drift, usually introducing many of false positives in the training dataset. To alleviate this problem, we propose in this paper a training protocol based on the self-paced learning paradigm. The main idea is to iteratively select a subset of images and boxes that are the most reliable, and use them for training. While in the past few years similar strategies have been adopted for SVMs and other classifiers, we are the first showing that a self-paced approach can be used with deep-network-based classifiers in an end-to-end training pipeline. The method we propose is built on the fully-supervised Fast-RCNN architecture and can be applied to similar architectures which represent the input image as a bag of boxes. We show state-of-the-art results on Pascal VOC 2007, Pascal VOC 2010 and ILSVRC 2013. OnILSVRC 2013 our results based on a low-capacity AlexNet network outperform even those weakly-supervised approaches which are based on much higher-capacity networks.