Àgata Lapedriza

Learning Deep Features for Discriminative Localization

In this work, we revisit the global average pooling layer proposed in [13], and shed light on how it explicitly enables the convolutional neural network (CNN) to have remarkable localization ability despite being trained on imagelevel labels. While this technique was previously proposed as a means for regularizing training, we find that it actually builds a generic localizable deep representation that exposes the implicit attention of CNNs on an image. Despite the apparent simplicity of global average pooling, we are able to achieve 37.1% top-5 error for object localization on ILSVRC 2014 without training on any bounding box annotation. We demonstrate in a variety of experiments that our network is able to localize the discriminative image regions despite just being trained for solving classification task1.

Robust gait-based gender classification using depth cameras

This article presents a new approach for gait-based gender recognition using depth cameras, that can run in real time. The main contribution of this study is a new fast feature extraction strategy that uses the 3D point cloud obtained from the frames in a gait cycle. For each frame, these points are aligned according to their centroid and grouped. After that, they are projected into their PCA plane, obtaining a representation of the cycle particularly robust against view changes. Then, final discriminative features are computed by first making a histogram of the projected points and then using linear discriminant analysis. To test the method we have used the DGait database, which is currently the only publicly available database for gait analysis that includes depth information. We have performed experiments on manually labeled cycles and over whole video sequences, and the results show that our method improves the accuracy significantly, compared with state-of-the-art systems which do not use depth information. Furthermore, our approach is insensitive to illumination changes, given that it discards the RGB information. That makes the method especially suitable for real applications, as illustrated in the last part of the experiments section.

Places: A 10 Million Image Database for Scene Recognition

The rise of multi-million-item dataset initiatives has enabled data-hungry machine learning algorithms to reach near-human semantic classification performance at tasks such as visual object and scene recognition. Here we describe the Places Database, a repository of 10 million scene photographs, labeled with scene semantic categories, comprising a large and diverse list of the types of environments encountered in the world. Using the state-of-the-art Convolutional Neural Networks (CNNs), we provide scene classification CNNs (Places-CNNs) as baselines, that significantly outperform the previous approaches. Visualization of the CNNs trained on Places shows that object detectors emerge as an intermediate representation of scene classification. With its high-coverage and high-diversity of exemplars, the Places Database along with the Places-CNNs offer a novel resource to guide future progress on scene recognition problems.

Places: An Image Database for Deep Scene Understanding

The rise of multi-million-item dataset initiatives has enabled data-hungry machine learning algorithms to reach near-human semantic classification at tasks such as object and scene recognition. Here we describe the Places Database, a repository of 10 million scene photographs, labeled with scene semantic categories and attributes, comprising a quasi-exhaustive list of the types of environments encountered in the world. Using state of the art Convolutional Neural Networks, we provide impressive baseline performances at scene classification. With its high-coverage and high-diversity of exemplars, the Places Database offers an ecosystem to guide future progress on currently intractable visual recognition problems.

Gender Recognition in Non Controlled Environments

In most of the automatic face classification applications, images should be captured in natural environments, where partial occlusions or high local changes in the illumination are frequent. For this reason, face classification tasks in uncontrolled environment are still nowadays unsolved problems, given that the loss of information caused by these artifacts can easily mislead any classifier. We present in this paper a system to extract robust face features that can be applied to encode information from any zone of the face and that can be used for different face classification problems. To test this method we include the results obtained in different gender classification experiments, considering controlled and uncontrolled environments and extracting face features from internal and external face zones. The obtained rates show, on the one hand, that we can obtain significant information applying the presented feature extraction scheme and, on the other hand, that the external face zone can contribute useful information for classification purposes

Boosted Online Learning for Face Recognition

Face recognition applications commonly suffer from three main drawbacks: a reduced training set, information lying in high-dimensional subspaces, and the need to incorporate new people to recognize. In the recent literature, the extension of a face classifier in order to include new people in the model has been solved using online feature extraction techniques. The most successful approaches of those are the extensions of the principal component analysis or the linear discriminant analysis. In the current paper, a new online boosting algorithm is introduced: a face recognition method that extends a boosting-based classifier by adding new classes while avoiding the need of retraining the classifier each time a new person joins the system. The classifier is learned using the multitask learning principle where multiple verification tasks are trained together sharing the same feature space. The new classes are added taking advantage of the structure learned previously, being the addition of new classes not computationally demanding. The present proposal has been (experimentally) validated with two different facial data sets by comparing our approach with the current state-of-the-art techniques. The results show that the proposed online boosting algorithm fares better in terms of final accuracy. In addition, the global performance does not decrease drastically even when the number of classes of the base problem is multiplied by eight.

Depth information in human gait analysis: an experimental study on gender recognition

This work presents DGait, a new gait database acquired with a depth camera. This database contains videos from 53 subjects walking in different directions. The intent of this database is to provide a public set to explore whether the depth can be used as an additional information source for gait classification purposes. Each video is labelled according to subject, gender and age. Furthermore, for each subject and view point, we provide initial and final frames of an entire walk cycle. On the other hand, we perform gait-based gender classification experiments with DGait database, in order to illustrate the usefulness of depth information for this purpose. In our experiments, we extract 2D and 3D gait features based on shape descriptors, and compare the performance of these features for gender identification, using a Kernel SVM. The obtained results show that depth can be an information source of great relevance for gait classification problems.

Preferred Spatial Frequencies for Human Face Processing Are Associated with Optimal Class Discrimination in the Machine

Psychophysical studies suggest that humans preferentially use a narrow band of low spatial frequencies for face recognition. Here we asked whether artificial face recognition systems have an improved recognition performance at the same spatial frequencies as humans. To this end, we estimated recognition performance over a large database of face images by computing three discriminability measures: Fisher Linear Discriminant Analysis, Non-Parametric Discriminant Analysis, and Mutual Information. In order to address frequency dependence, discriminabilities were measured as a function of (filtered) image size. All three measures revealed a maximum at the same image sizes, where the spatial frequency content corresponds to the psychophysical found frequencies. Our results therefore support the notion that the critical band of spatial frequencies for face recognition in humans and machines follows from inherent properties of face images, and that the use of these frequencies is associated with optimal face recognition performance.

Are External Face Features Useful for Automatic Face Classification

In this paper a new experiment using the FRGC database is proposed. The experiment deals with the use of external face features for face classification. Unlike the most part of algorithms that can be found in the literature for classifying faces, we consider the external information located at hair and ears as a reliable source of information. These features have often been discarded due to the difficulty of their extraction and alignment, and the lack of robustness in security related applications. Nevertheless, there are a lot of applications where these considerations are not valid, and the proper processing of external features can be an important additional source of information for classifications tasks. We also propose, following this assumption, a method for extracting external information from face images. The method is based on a top-down reconstructionbased algorithm for extracting the external face features. Once extracted, they are encoded in a second step using the Non Negative Matrix Factorization (NMF) algorithm, yielding an aligned high dimensional feature vector. This method has been used in a gender recognition problem, concluding that the encoded information is useful for classification purposes.

A sparse Bayesian approach for joint feature selection and classifier learning

In this paper we present a new method for Joint Feature Selection and Classifier Learning using a sparse Bayesian approach. These tasks are performed by optimizing a global loss function that includes a term associated with the empirical loss and another one representing a feature selection and regularization constraint on the parameters. To minimize this function we use a recently proposed technique, the Boosted Lasso algorithm, that follows the regularization path of the empirical risk associated with our loss function. We develop the algorithm for a well known non-parametrical classification method, the relevance vector machine, and perform experiments using a synthetic data set and three databases from the UCI Machine Learning Repository. The results show that our method is able to select the relevant features, increasing in some cases the classification accuracy when feature selection is performed.