Abstract: Background: COVID-19 infection, especially in cases with pneumonia, is associated with a high rate of pulmonary embolism (PE). In patients with contraindications for CT pulmonary angiography (CTPA) or non-diagnostic CTPA, perfusion single-photon emission computed tomography/computed tomography (Q-SPECT/CT) is a diagnostic alternative. The goal of this study is to develop a radiomic diagnostic system to detect PE based only on the analysis of Q-SPECT/CT scans.
Methods: This radiomic diagnostic system is based on a local analysis of Q-SPECT/CT volumes that includes both CT and Q-SPECT values for each volume point. We present a combined approach that uses radiomic features extracted from each scan as input into a fully connected classifcation neural network that optimizes a weighted crossentropy loss trained to discriminate between three diferent types of image patterns (pixel sample level): healthy lungs (control group), PE and pneumonia. Four types of models using diferent confguration of parameters were tested.
Results: The proposed radiomic diagnostic system was trained on 20 patients (4,927 sets of samples of three types of image patterns) and validated in a group of 39 patients (4,410 sets of samples of three types of image patterns). In the training group, COVID-19 infection corresponded to 45% of the cases and 51.28% in the test group. In the test group, the best model for determining diferent types of image patterns with PE presented a sensitivity, specifcity, positive predictive value and negative predictive value of 75.1%, 98.2%, 88.9% and 95.4%, respectively. The best model for detecting
pneumonia presented a sensitivity, specifcity, positive predictive value and negative predictive value of 94.1%, 93.6%, 85.2% and 97.6%, respectively. The area under the curve (AUC) was 0.92 for PE and 0.91 for pneumonia. When the results obtained at the pixel sample level are aggregated into regions of interest, the sensitivity of the PE increases to 85%, and all metrics improve for pneumonia.
Conclusion: This radiomic diagnostic system was able to identify the diferent lung imaging patterns and is a frst step toward a comprehensive intelligent radiomic system to optimize the diagnosis of PE by Q-SPECT/CT.

Abstract: Color induction is the phenomenon where the physical and the perceived colors of an object differ owing to the color distribution and the spatial configuration of the surrounding objects. Previous works studying this phenomenon on the lsY MacLeod–Boynton color space, show that color assimilation is present only when the magnocellular pathway (i.e., the Y axis) is activated (i.e., when there are luminance differences). Concretely, the authors showed that the effect is mainly induced by the koniocellular pathway (s axis), but not by the parvocellular pathway (l axis), suggesting that when magnocellular pathway is activated it inhibits the koniocellular pathway. In the present work, we study whether parvo-, konio-, and magnocellular pathways may influence on each other through the color induction effect. Our results show that color assimilation does not depend on a chromatic–chromatic interaction, and that chromatic assimilation is driven by the interaction between luminance and chromatic channels (mainly the magno- and the koniocellular pathways). Our results also show that chromatic induction is greatly decreased when all three visual pathways are simultaneously activated, and that chromatic pathways could influence each other through the magnocellular (luminance) pathway. In addition, we observe that chromatic channels can influence the luminance channel, hence inducing a small brightness induction. All these results show that color induction is a highly complex process where interactions between the several visual pathways are yet unknown and should be studied in greater detail.

Abstract: Continuous Hand Gesture Recognition (CHGR) has been extensively studied by researchers in the last few decades. Recently, one model has been presented to deal with the challenge of the boundary detection of isolated gestures in a continuous gesture video [17]. To enhance the model performance and also replace the handcrafted feature extractor in the presented model in [17], we propose a GCN model and combine it with the stacked Bi-LSTM and Attention modules to push the temporal information in the video stream. Considering the breakthroughs of GCN models for skeleton modality, we propose a two-layer GCN model to empower the 3D hand skeleton features. Finally, the class probabilities of each isolated gesture are fed to the post-processing module, borrowed from [17]. Furthermore, we replace the anatomical graph structure with some non-anatomical graph structures. Due to the lack of a large dataset, including both the continuous gesture sequences and the corresponding isolated gestures, three public datasets in Dynamic Hand Gesture Recognition (DHGR), RKS-PERSIANSIGN, and ASLVID, are used for evaluation. Experimental results show the superiority of the proposed model in dealing with isolated gesture boundaries detection in continuous gesture sequences

Abstract: Data-driven paradigms using machine learning are becoming ubiquitous in image processing and communications. In particular, image-to-image (I2I) translation is a generic and widely used approach to image processing problems, such as image synthesis, style transfer, and image restoration. At the same time, neural image compression has emerged as a data-driven alternative to traditional coding approaches in visual communications. In this paper, we study the combination of these two paradigms into a joint I2I compression and translation framework, focusing on multi-domain image synthesis. We first propose distributed I2I translation by integrating quantization and entropy coding into an I2I translation framework (i.e. I2Icodec). In practice, the image compression functionality (i.e. autoencoding) is also desirable, requiring to deploy alongside I2Icodec a regular image codec. Thus, we further propose a unified framework that allows both translation and autoencoding capabilities in a single codec. Adaptive residual blocks conditioned on the translation/compression mode provide flexible adaptation to the desired functionality. The experiments show promising results in both I2I translation and image compression using a single model.

Abstract: Deep learning has drastically changed computer vision in the past decades and achieved great success in many applications, such as image classification, retrieval, detection, and segmentation thanks to the emergence of neural networks. Typically, for most applications, these networks are presented with examples from all tasks they are expected to perform. However, for many applications, this is not a realistic
scenario, and an algorithm is required to learn tasks sequentially. Continual learning proposes theory and methods for this scenario.
The main challenge for continual learning systems is called catastrophic forgetting and refers to a significant drop in performance on previous tasks. To tackle this problem, three main branches of methods have been explored to alleviate the forgetting in continual learning. They are regularization-based methods, rehearsalbased methods, and parameter isolation-based methods. However, most of them are focused on image classification tasks. Continual learning of many computer vision fields has still not been well-explored. Thus, in this thesis, we extend the continual learning knowledge to meta learning, we propose a method for the incremental learning of hierarchical relations for image classification, we explore image recognition in online continual learning, and study continual learning for cross-modal learning.
In this thesis, we explore the usage of image rehearsal when addressing the incremental meta learning problem. Observing that existingmethods fail to improve performance with saved exemplars, we propose to mix exemplars with current task data and episode-level distillation to overcome forgetting in incremental meta learning. Next, we study a more realistic image classification scenario where each class has multiple granularity levels. Only one label is present at any time, which requires the model to infer if the provided label has a hierarchical relation with any already known label. In experiments, we show that the estimated hierarchy information can be beneficial in both the training and inference stage.
For the online continual learning setting, we investigate the usage of intermediate feature replay. In this case, the training samples are only observed by the model only one time. Here we fix thememory buffer for feature replay and compare the effectiveness of saving features from different layers. Finally, we investigate multi-modal continual learning, where an image encoder is cooperating with a semantic branch. We consider the continual learning of both zero-shot learning and cross-modal retrieval problems.

Abstract: Despite decades of research in Optical Music Recognition (OMR), the recognition of old handwritten music scores remains a challenge because of the variabilities in the handwriting styles, paper degradation, lack of standard notation, etc. Therefore, the research in OMR systems adapted to the particularities of old manuscripts is crucial to accelerate the conversion of music scores existing in archives into digital libraries, fostering the dissemination and preservation of our music heritage. In this paper we explore the adaptation of sequence-to-sequence models with attention mechanism (used in translation and handwritten text recognition) and the generation of specific synthetic data for recognizing old music scores. The experimental validation demonstrates that our approach is promising, especially when compared with long short-term memory neural networks.

Abstract: Despite recent advances in 3D pose estimation of human hands, especially thanks to the advent of CNNs and depth cameras, this task is still far from being solved. This is mainly due to the highly non-linear dynamics of fingers, which make hand model training a challenging task. In this paper, we exploit a novel hierarchical tree-like structured CNN, in which branches are trained to become specialized in predefined subsets of hand joints, called local poses. We further fuse local pose features, extracted from hierarchical CNN branches, to learn higher order dependencies among joints in the final pose by end-to-end training. Lastly, the loss function used is also defined to incorporate appearance and physical constraints about doable hand motion and deformation. Finally, we introduce a non-rigid data augmentation approach to increase the amount of training depth data. Experimental results suggest that feeding a tree-shaped CNN, specialized in local poses, into a fusion network for modeling joints correlations and dependencies, helps to increase the precision of final estimations, outperforming state-of-the-art results on NYU and SyntheticHand datasets.