Inspired by the conclusion that human choose the visual cortex regions which corresponding to the real size of the object to analyze the features of the object, when realizing the objects in the real world. This paper presents a framework -- SizeNet which based on both the real sizes and the features of objects, to solve objects recognition problems. SizeNet was used for the objects recognition experiments on the homemade Rsize dataset, and compared with State-of-the-art Methods AlexNet, VGG-16, Inception V3, Resnet-18 DenseNet-121. The results show that SizeNet provides much higher accuracy rates for the objects recognition than the other algorithms. SizeNet can solve the two problems that correctly recognize the objects whose features are highly similar but the real sizes are obviously different from each other, and correctly distinguish the target object from the interference objects whose real sizes are obviously different from the target object. This is because SizeNet recognizes the object based not only the features, but also the real size. The real size of the object can help to exclude the interference object categories whose real size ranges do not match the real size of the object, which greatly reducing the object categories' number in the label set used for the downstream object recognition based on object features. SizeNet is of great significance to the study of interpretable computer vision. Our code and dataset will be made public.

Most successful computer vision models transform low-level features, such as Gabor filter responses, into richer representations of intermediate or mid-level complexity for downstream visual tasks. These mid-level representations have not been explored for event cameras, although it is especially relevant to the visually sparse and often disjoint spatial information in the event stream. By making use of locally consistent intermediate representations, termed as superevents, numerous visual tasks ranging from semantic segmentation, visual tracking, depth estimation shall benefit. In essence, superevents are perceptually consistent local units that delineate parts of an object in a scene. Inspired by recent deep learning architectures, we present a novel method that employs lifetime augmentation for obtaining an event stream representation that is fed to a fully convolutional network to extract superevents. Our qualitative and quantitative experimental results on several sequences of a benchmark dataset highlights the significant potential for event-based downstream applications.

Dialogue systems are a popular Natural Language Processing (NLP) task as it is promising in real-life applications. It is also a complicated task since many NLP tasks deserving study are involved. As a result, a multitude of novel works on this task are carried out, and most of them are deep learning-based due to the outstanding performance. In this survey, we mainly focus on the deep learning-based dialogue systems. We comprehensively review state-of-the-art research outcomes in dialogue systems and analyze them from two angles: model type and system type. Specifically, from the angle of model type, we discuss the principles, characteristics, and applications of different models that are widely used in dialogue systems. This will help researchers acquaint these models and see how they are applied in state-of-the-art frameworks, which is rather helpful when designing a new dialogue system. From the angle of system type, we discuss task-oriented and open-domain dialogue systems as two streams of research, providing insight into the hot topics related. Furthermore, we comprehensively review the evaluation methods and datasets for dialogue systems to pave the way for future research. Finally, some possible research trends are identified based on the recent research outcomes. To the best of our knowledge, this survey is the most comprehensive and up-to-date one at present in the area of dialogue systems and dialogue-related tasks, extensively covering the popular frameworks, topics, and datasets. Keywords: Dialogue Systems, Chatbots, Conversational AI, Task-oriented, Open Domain, Chit-chat, Question Answering, Artificial Intelligence, Natural Language Processing, Information Retrieval, Deep Learning, Neural Networks, CNN, RNN, Hierarchical Recurrent Encoder-Decoder, Memory Networks, Attention, Transformer, Pointer Net, CopyNet, Reinforcement Learning, GANs, Knowledge Graph, Survey, Review

Neural networks (NNs) have shown high predictive performance, however, with shortcomings. Firstly, the reasons behind the classifications are not fully understood. Several explanation methods have been developed, but they do not provide mechanisms for users to interact with the explanations. Explanations are social, meaning they are a transfer of knowledge through interactions. Nonetheless, current explanation methods contribute only to one-way communication. Secondly, NNs tend to be overconfident, providing unreasonable uncertainty estimates on out-of-distribution observations. We overcome these difficulties by training a Bayesian convolutional neural network (CNN) that uses explanation feedback. After training, the model presents explanations of training sample classifications to an annotator. Based on the provided information, the annotator can accept or reject the explanations by providing feedback. Our proposed method utilizes this feedback for fine-tuning to correct the model such that the explanations and classifications improve. We use existing CNN architectures to demonstrate the method's effectiveness on one toy dataset (decoy MNIST) and two real-world datasets (Dogs vs. Cats and ISIC skin cancer). The experiments indicate that few annotated explanations and fine-tuning epochs are needed to improve the model and predictive performance, making the model more trustworthy and understandable.

Unlike areas such as computer vision and speech recognition where convolutional and recurrent neural networks-based approaches have proven effective to the nature of the respective areas of application, deep learning (DL) still lacks a general approach suitable for the unique nature and challenges of RF systems such as radar, signals intelligence, electronic warfare, and communications. Existing approaches face problems in robustness, consistency, efficiency, repeatability and scalability. One of the main challenges in RF sensing such as radar target identification is the difficulty and cost of obtaining data. Hundreds to thousands of samples per class are typically used when training for classifying signals into 2 to 12 classes with reported accuracy ranging from 87% to 99%, where accuracy generally decreases with more classes added. In this paper, we present a new DL approach based on multistage training and demonstrate it on RF sensing signal classification. We consistently achieve over 99% accuracy for up to 17 diverse classes using only 11 samples per class for training, yielding up to 35% improvement in accuracy over standard DL approaches.

The assignment of importance scores to particular frames or (short) segments in a video is crucial for summarization, but also a difficult task. Previous work utilizes only one source of visual features. In this paper, we suggest a novel model architecture that combines three feature sets for visual content and motion to predict importance scores. The proposed architecture utilizes an attention mechanism before fusing motion features and features representing the (static) visual content, i.e., derived from an image classification model. Comprehensive experimental evaluations are reported for two well-known datasets, SumMe and TVSum. In this context, we identify methodological issues on how previous work used these benchmark datasets, and present a fair evaluation scheme with appropriate data splits that can be used in future work. When using static and motion features with parallel attention mechanism, we improve state-of-the-art results for SumMe, while being on par with the state of the art for the other dataset.

This paper presents a new network intrusion detection system (NIDS) based on Graph Neural Networks (GNNs). GNNs are a relatively new sub-field of deep neural networks, which have the unique ability to leverage the inherent structure of graph-based data. Training and evaluation data for NIDSs are typically represented as flow records, which can naturally be represented in a graph format. This establishes the potential and motivation for exploring GNNs for the purpose of network intrusion detection, which is the focus of this paper. E-GraphSAGE, our proposed new approach is based on the established GraphSAGE model, but provides the necessary modifications in order to support edge features for edge classification, and hence the classification of network flows into benign and attack classes. An extensive experimental evaluation based on six recent NIDS benchmark datasets shows the excellent performance of our E-GraphSAGE based NIDS in comparison with the state-of-the-art.

Supervised deep learning has reached the level of radiologists in 11 out of 14 pathologies. Supervised learning is an automatic learning task consisting in learning a prediction function from annotated examples. For the detection of pleural effusion, AI does as well as the radiologist, but this is not the case for the diagnosis of thoracic hernia. For Pleural Effusion, the annotated data set is 11000 examples compared to 110 examples for hernia. The number of correctly annotated data is one of the determinants of the effectiveness of supervised learning.

The biggest hurdle to clear when developing machine learning solutions has always been the data. Large-scale, clean, fully-annotated data sets like ImageNet and COCO are not easy to come by, especially for niche tasks. This is especially true of deep learning and will become more true as deep learning is increasingly adopted. To overcome the labeled data bottleneck, researchers and developers are developing a variety of techniques like transfer learning, domain adaptation, synthetic data generation, and many semi- and self-supervised techniques. The world of semi- and self-supervised techniques is a particularly fascinating field as it seems almost like magic.

To best assist human users while they complete everyday tasks, robots should be able to understand their queries, answer them and perform actions accordingly. In other words, they should be able to flexibly generate and perform actions that are aligned with a user's verbal instructions. To understand a user's instructions and act accordingly, robotic systems should be able to make associations between linguistic expressions, actions and environments. Deep neural networks have proved to be particularly good at acquiring representations of linguistic expressions, yet they typically need to be trained on large datasets including robot actions, linguistic descriptions and information about different environments. Researchers at Waseda University in Tokyo recently developed a deep neural network that can acquire grounded representations of robot actions and linguistic descriptions of these actions.