Abstract: This paper presents the combined use of gradient and mutual information for infrared and intensity templates matching. We propose to joint: (i) feature matching in a multiresolution context and (ii) information propagation through scale-space representations. Our method consists in combining mutual information with a shape descriptor based on gradient, and propagate them following a coarse-to-fine strategy. The main contributions of this work are: to offer a theoretical formulation towards a multimodal stereo matching; to show that gradient and mutual information can be reinforced while they are propagated between consecutive levels; and to show that they are valid cost functions in multimodal template matchings. Comparisons are presented showing the improvements and viability of the proposed approach.

Abstract: In object-class segmentation, one of the most tedious tasks is to manually segment many object examples in order to learn a model of the object category. Yet, there has been little research on reducing the degree of manual annotation for
object-class segmentation. In this work we explore alternative strategies which do not require full manual segmentation of the object in the training set. In particular, we study the use of bounding boxes as a coarser and much cheaper form of segmentation and we perform a comparative study of several Multiple-Instance Learning techniques that allow to obtain a model with this type of weak annotation. We show that some of these methods can be competitive, when used with coarse
segmentations, with methods that require full manual segmentation of the objects. Furthermore, we show how to use active learning combined with this weakly supervised strategy.
As we see, this strategy permits to reduce the amount of annotation and optimize the number of examples that require full manual segmentation in the training set.

Abstract: Deep learning for human action recognition in videos is making significant progress, but is slowed down by its dependency on expensive manual labeling of large video collections. In this work, we investigate the generation of synthetic training data for action recognition, as it has recently shown promising results for a variety of other computer vision tasks. We propose an interpretable parametric generative model of human action videos that relies on procedural generation and other computer graphics techniques of modern game engines. We generate a diverse, realistic, and physically plausible dataset of human action videos, called PHAV for ”Procedural Human Action Videos”. It contains a total of 39, 982 videos, with more than 1, 000 examples for each action of 35 categories. Our approach is not limited to existing motion capture sequences, and we procedurally define 14 synthetic actions. We introduce a deep multi-task representation learning architecture to mix synthetic and real videos, even if the action categories differ. Our experiments on the UCF101 and HMDB51 benchmarks suggest that combining our large set of synthetic videos with small real-world datasets can boost recognition performance, significantly
outperforming fine-tuning state-of-the-art unsupervised generative models of videos.

Abstract: State-of-the-art approaches for semantic image segmentation are built on Convolutional Neural Networks (CNNs). The typical segmentation architecture is composed of (a) a downsampling path responsible for extracting coarse semantic features, followed by (b) an upsampling path trained to recover the input image resolution at the output of the model and, optionally, (c) a post-processing module (e.g. Conditional Random Fields) to refine the model predictions.
Recently, a new CNN architecture, Densely Connected Convolutional Networks (DenseNets), has shown excellent results on image classification tasks. The idea of DenseNets is based on the observation that if each layer is directly connected to every other layer in a feed-forward fashion then the network will be more accurate and easier to train.
In this paper, we extend DenseNets to deal with the problem of semantic segmentation. We achieve state-of-the-art results on urban scene benchmark datasets such as CamVid and Gatech, without any further post-processing module nor pretraining. Moreover, due to smart construction of the model, our approach has much less parameters than currently published best entries for these datasets.

Abstract: Wearable cameras are become more popular in recent years for capturing the unscripted moments of the first-person that help to analyze the users lifestyle. In this work, we aim to recognize the places related to food in egocentric images during a day to identify the daily food patterns of the first-person. Thus, this system can assist to improve their eating behavior to protect users against food-related diseases. In this paper, we use Siamese Neural Networks to learn the similarity between images from corresponding inputs for one-shot food places classification. We tested our proposed method with ‘MiniEgoFoodPlaces’ with 15 food related places. The proposed Siamese Neural Networks model with MobileNet achieved an overall classification accuracy of 76.74% and 77.53% on the validation and test sets of the “MiniEgoFoodPlaces” dataset, respectively outperforming with the base models, such as ResNet50, InceptionV3, and InceptionResNetV2.

Abstract: Domain Adaptation methods can be classified into two basic families of approaches: non-parametric and parametric. Non-parametric approaches depend on statistical indicators such as feature covariances to minimize the domain shift. Non-parametric approaches tend to be fast to compute and require no additional parameters, but they are unable to leverage probability density functions with complex internal structures. Parametric approaches, on the other hand, use models of the probability distributions as surrogates in minimizing the domain shift, but they require additional trainable parameters to model these distributions. In this work, we propose a new statistical approach to minimizing the domain shift based on stochastically projecting and evaluating the cumulative density function in both domains. As with non-parametric approaches, there are no additional trainable parameters. As with parametric approaches, the internal structure of both domains’ probability distributions is considered, thus leveraging a higher amount of information when reducing the domain shift. Evaluation on standard datasets used for Domain Adaptation shows better performance of the proposed model compared to non-parametric approaches while being competitive with parametric ones. (Code available at: https://github.com/moliusimon/mca).