Elisa Fromont

Discriminative feature fusion for image classification

Bag-of-words-based image classification approaches mostly rely on low level local shape features. However, it has been shown that combining multiple cues such as color, texture, or shape is a challenging and promising task which can improve the classification accuracy. Most of the state-of-the-art feature fusion methods usually aim to weight the cues without considering their statistical dependence in the application at hand. In this paper, we present a new logistic regression-based fusion method, called LRFF, which takes advantage of the different cues without being tied to any of them. We also design a new marginalized kernel by making use of the output of the regression model. We show that such kernels, surprisingly ignored so far by the computer vision community, are particularly well suited to achieve image classification tasks. We compare our approach with existing methods that combine color and shape on three datasets. The proposed learning-based feature fusion process clearly outperforms the state-of-the art fusion methods for image classification.

Mining Mid-level Features for Image Classification

Mid-level or semi-local features learnt using class-level information are potentially more distinctive than the traditional low-level local features constructed in a purely bottom-up fashion. At the same time they preserve some of the robustness properties with respect to occlusions and image clutter. In this paper we propose a new and effective scheme for extracting mid-level features for image classification, based on relevant pattern mining. In particular, we mine relevant patterns of local compositions of densely sampled low-level features. We refer to the new set of obtained patterns as Frequent Local Histograms or FLHs. During this process, we pay special attention to keeping all the local histogram information and to selecting the most relevant reduced set of FLH patterns for classification. The careful choice of the visual primitives and an extension to exploit both local and global spatial information allow us to build powerful bag-of-FLH-based image representations. We show that these bag-of-FLHs are more discriminative than traditional bag-of-words and yield state-of-the-art results on various image classification benchmarks, including Pascal VOC.

Supervised learning of Gaussian mixture models for visual vocabulary generation

The creation of semantically relevant clusters is vital in bag-of-visual words models which are known to be very successful to achieve image classification tasks. Generally, unsupervised clustering algorithms, such as K-means, are employed to create such clusters from which visual dictionaries are deduced. K-means achieves a hard assignment by associating each image descriptor to the cluster with the nearest mean. By this way, the within-cluster sum of squares of distances is minimized. A limitation of this approach in the context of image classification is that it usually does not use any supervision that limits the discriminative power of the resulting visual words (typically the centroids of the clusters). More recently, some supervised dictionary creation methods based on both supervised information and data fitting were proposed leading to more discriminative visual words. But, none of them consider the uncertainty present at both image descriptor and cluster levels. In this paper, we propose a supervised learning algorithm based on a Gaussian mixture model which not only generalizes the K-means algorithm by allowing soft assignments, but also exploits supervised information to improve the discriminative power of the clusters. Technically, our algorithm aims at optimizing, using an EM-based approach, a convex combination of two criteria: the first one is unsupervised and based on the likelihood of the training data; the second is supervised and takes into account the purity of the clusters. We show on two well-known datasets that our method is able to create more relevant clusters by comparing its behavior with the state of the art dictionary creation methods.

A computerized prediction model of hazardous inflammatory platelet transfusion outcomes.

Background Platelet component (PC) transfusion leads occasionally to inflammatory hazards. Certain BRMs that are secreted by the platelets themselves during storage may have some responsibility. Methodology/Principal Findings First, we identified non-stochastic arrangements of platelet-secreted BRMs in platelet components that led to acute transfusion reactions (ATRs). These data provide formal clinical evidence that platelets generate secretion profiles under both sterile activation and pathological conditions. We next aimed to predict the risk of hazardous outcomes by establishing statistical models based on the associations of BRMs within the incriminated platelet components and using decision trees. We investigated a large (n = 65) series of ATRs after platelet component transfusions reported through a very homogenous system at one university hospital. Herein, we used a combination of clinical observations, ex vivo and in vitro investigations, and mathematical modeling systems. We calculated the statistical association of a large variety (n = 17) of cytokines, chemokines, and physiologically likely factors with acute inflammatory potential in patients presenting with severe hazards. We then generated an accident prediction model that proved to be dependent on the level (amount) of a given cytokine-like platelet product within the indicated component, e.g., soluble CD40-ligand (>289.5 pg/109 platelets), or the presence of another secreted factor (IL-13, >0). We further modeled the risk of the patient presenting either a febrile non-hemolytic transfusion reaction or an atypical allergic transfusion reaction, depending on the amount of the chemokine MIP-1α (<20.4 or >20.4 pg/109 platelets, respectively). Conclusions/Significance This allows the modeling of a policy of risk prevention for severe inflammatory outcomes in PC transfusion.

Platelet components associated with adverse reactions: predictive value of mitochondrial DNA relative to biological response modifiers

Biological response modifiers (BRMs), secreted by platelets (PLTs) during storage, play a role in adverse events (AEs) associated with transfusion. Moreover, mitochondrial DNA (mtDNA) levels in PLT components (PCs) are associated with AEs. In this study we explore whether there is a correlation between pathogenic BRMs and mtDNA levels and whether these markers can be considered predictors of transfusion pathology.

Optimal constraint-based decision tree induction from itemset lattices

In this article we show that there is a strong connection between decision tree learning and local pattern mining. This connection allows us to solve the computationally hard problem of finding optimal decision trees in a wide range of applications by post-processing a set of patterns: we use local patterns to construct a global model. We exploit the connection between constraints in pattern mining and constraints in decision tree induction to develop a framework for categorizing decision tree mining constraints. This framework allows us to determine which model constraints can be pushed deeply into the pattern mining process, and allows us to improve the state-of-the-art of optimal decision tree induction.

Mining spatiotemporal patterns in dynamic plane graphs

Dynamic graph mining is the task of searching for subgraph patterns that capture the evolution of a dynamic graph. In this paper, we are interested in mining dynamic graphs in videos. A video can be regarded as a dynamic graph, whose evolution over time is represented by a series of plane graphs, one graph for each video frame. As such, subgraph patterns in this series may correspond to objects that frequently appear in the video. Furthermore, by associating spatial information to each of the nodes in these graphs, it becomes possible to track a given object through the video in question. We present, in this paper, two plane graph mining algorithms, called plagram and dyplagram, for the extraction of spatiotemporal patterns. A spatiotemporal pattern is a set of occurrences of a given subgraph pattern which are not too far apart w.r.t time nor space. Experiments demonstrate that our algorithms are effective even in contexts where general-purpose algorithms would not provide the complete set of frequent subgraphs. We also show that they give promising results when applied to object tracking in videos.

Multi-task, multi-domain learning

We present an approach that leverages multiple datasets annotated for different tasks (e.g., classification with different labelsets) to improve the predictive accuracy on each individual dataset. Domain adaptation techniques can correct dataset bias but they are not applicable when the tasks differ, and they need to be complemented to handle multi-task settings. We propose a new selective loss function that can be integrated into deep neural networks to exploit training data coming from multiple datasets annotated for related but possibly different labelsets. We show that the gradient-reversal approach for domain adaptation can be used in this setup to additionally handle domain shifts. We also propose an auto-context approach that further captures existing correlations across tasks. Thorough experiments on two types of applications (semantic segmentation and hand pose estimation) show the relevance of our approach in different contexts.

Platelet soluble CD40-ligand level is associated with transfusion adverse reactions in a mixed threshold-and-hit model

To the editor: Platelets are the principal source of soluble CD40-ligand (sCD40L) found in blood.[1][1] This biological response modifier has been reported to be a candidate mediator for acute reactions after platelet transfusions.[2][2] Serious adverse reactions (SARs) associated with excessive

Scene classification based on semantic labeling

Finding an appropriate image representation is a crucial problem in robotics. This problem has been classically addressed by means of computer vision techniques, where local and global features are used. The selection or/and combination of different features is carried out by taking into account repeatability and distinctiveness, but also the specific problem to solve. In this article, we propose the generation of image descriptors from general purpose semantic annotations. This approach has been evaluated as source of information for a scene classifier, and specifically using Clarifai as the semantic annotation tool. The experimentation has been carried out using the ViDRILO toolbox as benchmark, which includes a comparison of state-of-the-art global features and tools to make comparisons among them. According to the experimental results, the proposed descriptor performs similarly to well-known domain-specific image descriptors based on global features in a scene classification task. Moreover, the proposed descriptor is based on generalist annotations without any type of problem-oriented parameter tuning.