Abstract: Road detection is a vital task for the development of autonomous vehicles. The knowledge of the free road surface ahead of the target vehicle can be used for autonomous driving, road departure warning, as well as to support advanced driver assistance systems like vehicle or pedestrian detection. Using vision to detect the road has several advantages in front of other sensors: richness of features, easy integration, low cost or low power consumption. Common vision-based road detection approaches use low-level features (such as color or texture) as visual cues to group pixels exhibiting similar properties. However, it is difficult to foresee a perfect clustering algorithm since roads are in outdoor scenarios being imaged from a mobile platform. In this paper, we propose a novel high-level approach to vision-based road detection based on geographical information. The key idea of the algorithm is exploiting geographical information to provide a rough detection of the road. Then, this segmentation is refined at low-level using color information to provide the final result. The results presented show the validity of our approach.

Abstract: We propose an online domain adaptation method for the deformable part-based model (DPM). The online domain adaptation is based on a two-level hierarchical adaptation tree, which consists of instance detectors in the leaf nodes and a category detector at the root node. Moreover, combined with a multiple object tracking procedure (MOT), our proposal neither requires target-domain annotated data nor revisiting the source-domain data for performing the source-to-target domain adaptation of the DPM. From a practical point of view this means that, given a source-domain DPM and new video for training on a new domain without object annotations, our procedure outputs a new DPM adapted to the domain represented by the video. As proof-of-concept we apply our proposal to the challenging task of pedestrian detection. In this case, each instance detector is an exemplar classifier trained online with only one pedestrian per frame. The pedestrian instances are collected by MOT and the hierarchical model is constructed dynamically according to the pedestrian trajectories. Our experimental results show that the adapted detector achieves the accuracy of recent supervised domain adaptation methods (i.e., requiring manually annotated targetdomain data), and improves the source detector more than 10 percentage points.

Abstract: The Stixel World is a medium-level, compact representation of road scenes that abstracts millions of disparity pixels into hundreds or thousands of stixels. The goal of this work is to implement and evaluate a complete multi-stixel estimation pipeline on an embedded, energyefficient, GPU-accelerated device. This work presents a full GPU-accelerated implementation of stixel estimation that produces reliable results at 26 frames per second (real-time) on the Tegra X1 for disparity images of 1024×440 pixels and stixel widths of 5 pixels, and achieves more than 400 frames per second on a high-end Titan X GPU card.

Abstract: The increase in awareness of people towards their nutritional habits has drawn considerable attention to the field of automatic food analysis. Focusing on self-service restaurants environment, automatic food analysis is not only useful for extracting nutritional information from foods selected by customers, it is also of high interest to speed up the service solving the bottleneck produced at the cashiers in times of high demand. In this paper, we address the problem of automatic food tray analysis in canteens and restaurants environment, which consists in predicting multiple foods placed on a tray image. We propose a new approach for food analysis based on convolutional neural networks, we name Semantic Food Detection, which integrates in the same framework food localization, recognition and segmentation. We demonstrate that our method improves the state of the art food detection by a considerable margin on the public dataset UNIMIB2016 achieving about 90% in terms of F-measure, and thus provides a significant technological advance towards the automatic billing in restaurant environments.

Abstract: Deep convolutional networks can achieve impressive results on RGB scene recognition thanks to large data sets such as places. In contrast, RGB-D scene recognition is still underdeveloped in comparison, due to two limitations of RGB-D data we address in this paper. The first limitation is the lack of depth data for training deep learning models. Rather than fine tuning or transferring RGB-specific features, we address this limitation by proposing an architecture and a two-step training approach that directly learns effective depth-specific features using weak supervision via patches. The resulting RGB-D model also benefits from more complementary multimodal features. Another limitation is the short range of depth sensors (typically 0.5 m to 5.5 m), resulting in depth images not capturing distant objects in the scenes that RGB images can. We show that this limitation can be addressed by using RGB-D videos, where more comprehensive depth information is accumulated as the camera travels across the scenes. Focusing on this scenario, we introduce the ISIA RGB-D video data set to evaluate RGB-D scene recognition with videos. Our video recognition architecture combines convolutional and recurrent neural networks that are trained in three steps with increasingly complex data to learn effective features (i.e., patches, frames, and sequences). Our approach obtains the state-of-the-art performances on RGB-D image (NYUD2 and SUN RGB-D) and video (ISIA RGB-D) scene recognition.

Abstract: Humans modify their facial expressions in order to communicate their internal states and sometimes to mislead observers regarding their true emotional states. Evidence in experimental psychology shows that discriminative facial responses are short and subtle. This suggests that such behavior would be easier to distinguish when captured in high resolution at an increased frame rate. We are proposing SASE-FE, the first dataset of facial expressions that are either congruent or incongruent with underlying emotion states. We show that overall the problem of recognizing whether facial movements are expressions of authentic emotions or not can be successfully addressed by learning spatio-temporal representations of the data. For this purpose, we propose a method that aggregates features along fiducial trajectories in a deeply learnt space. Performance of the proposed model shows that on average, it is easier to distinguish among genuine facial expressions of emotion than among unfelt facial expressions of emotion and that certain emotion pairs such as contempt and disgust are more difficult to distinguish than the rest. Furthermore, the proposed methodology improves state of the art results on CK+ and OULU-CASIA datasets for video emotion recognition, and achieves competitive results when classifying facial action units on BP4D datase.

Abstract: Recent studies have shown that the environment where people eat can affect their nutritional behaviour. In this work, we provide automatic tools for a personalised analysis of a person's health habits by the examination of daily recorded egocentric photo-streams. Specifically, we propose a new automatic approach for the classification of food-related environments, that is able to classify up to 15 such scenes. In this way, people can monitor the context around their food intake in order to get an objective insight into their daily eating routine. We propose a model that classifies food-related scenes organized in a semantic hierarchy. Additionally, we present and make available a new egocentric dataset composed of more than 33000 images recorded by a wearable camera, over which our proposed model has been tested. Our approach obtains an accuracy and F-score of 56\% and 65\%, respectively, clearly outperforming the baseline methods.

Abstract: While many individual tasks in the domain of human analysis have recently received an accuracy boost from deep learning approaches, multi-task learning has mostly been ignored due to a lack of data. New synthetic datasets are being released, filling this gap with synthetic generated data. In this work, we analyze four related human analysis tasks in still images in a multi-task scenario by leveraging such datasets. Specifically, we study the correlation of 2D/3D pose estimation, body part segmentation and full-body depth estimation. These tasks are learned via the well-known Stacked Hourglass module such that each of the task-specific streams shares information with the others. The main goal is to analyze how training together these four related tasks can benefit each individual task for a better generalization. Results on the newly released SURREAL dataset show that all four tasks benefit from the multi-task approach, but with different combinations of tasks: while combining all four tasks improves 2D pose estimation the most, 2D pose improves neither 3D pose nor full-body depth estimation. On the other hand 2D parts segmentation can benefit from 2D pose but not from 3D pose. In all cases, as expected, the maximum improvement is achieved on those human body parts that show more variability in terms of spatial distribution, appearance and shape, e.g. wrists and ankles.