Search

Human-robot-collaboration requires robot to proactively and intelligently recognize the intention of human operator. Despite deep learning approaches have achieved certain results in performing feature learning and long-term temporal dependencies modeling, the motion prediction is still not desirable enough, which unavoidably compromises the accomplishment of tasks. Therefore, a hybrid recurrent neural network architecture is proposed for intention recognition to conduct the assembly tasks cooperatively. Specifically, the improved LSTM (ILSTM) and improved Bi-LSTM (IBi-LSTM) networks are first explored with state activation function and gate activation function to improve the network performance. The employment of the IBi-LSTM unit in the first layers of the hybrid architecture helps to learn the features effectively and fully from complex sequential data, and the LSTM-based cell in the last layer contributes to capturing the forward dependency. This hybrid network architecture can improve the prediction performance of intention recognition effectively. One experimental platform with the UR5 collaborative robot and human motion capture device is set up to test the performance of the proposed method. One filter, that is, the quartile-based amplitude limiting algorithm in sliding window, is designed to deal with the abnormal data of the spatiotemporal data, and thus, to improve the accuracy of network training and testing. The experimental results show that the hybrid network can predict the motion of human operator more precisely in collaborative workspace, compared with some representative deep learning methods.

In recent years, voice interaction-based control systems have attracted considerable attention for industrial control systems implementing Industrial Internet of Things (IIoT) technologies. The development of automated semantic understanding relates to the industrial Internet equipment used to realize remote voice control as well as to its intelligent management and control. In these emerging voice interaction-enabled industrial central control systems, sorting technologies are considered critical. For complex user questions, the level of satisfaction regarding the answers given by such systems tends to be low. Driven by these challenges and opportunities, the optimization of conventional retrieval-based question answering through deep learning methods has become popular. In this study, we propose three deep semantic sorting models based on deep learning, including a multi-layer convolutional matching sorting model for single documents and two interactive pairwise BERT(Bidirectional Encoder Representations from Transformers) sorting models for document pairs. Two main network architectures are proposed to model document pairs, named Pairwise-Twin-BERT and Pairwise-Triple-BERT. Experimental results indicate that proposed models performed better than state-of-the-art methods based on text matching in a candidate document sorting task.

Capturing the dynamics of facial expression progression in video is an essential and challenging task for facial expression recognition (FER). In this paper, we propose an effective framework to address this challenge. We develop a C3D-based network architecture, 3D-Inception-ResNet, to extract spatial-temporal features from the dynamic facial expression image sequence. A Spatial-Temporal and Channel Attention Module (STCAM) is proposed to explicitly exploit the holistic spatial-temporal and channel-wise correlations among the extracted features. Specifically, the proposed STCAM calculates a channel-wise and a spatial-temporal-wise attention map to enhance the features along the corresponding feature dimensions for more representative features. We evaluate our method on three popular dynamic facial expression recognition datasets, CK+, Oulu-CASIA and MMI. Experimental results show that our method achieves better or comparable performance compared to the state-of-the-art approaches.

With the emergence of various techniques involved in deep learning the researchers of computer vision tends to focus on the strategies such as object recognition and segmentation of image. This has inclined to accomplish the deep learning techniques in 3D reconstruction of both specific and generic objects. As the space for reconstruction of 3D images either in single or multi view has envisioned the researchers to concentrate on the available technologies used for reconstruction. With the available built in methods and technologies in deep learning, the performance of the proposed methods were reviewed and analyzed using several parameters. As the remaking of 2D images is still in the beginning stage, it is important to study the 3D shape representations, various network architecture, methodologies and approaches behind 3D reconstruction. In this work a review of deep learning methods for single or multiple RGB images of specific and generic object 3D reconstruction was done. Several methods and their importance were also discussed along with the challenges encountered and with further research directions. This paper critically analyzes the various 3D Shaped Representations, 3D Data Network Architectures, Depth Estimation methods, Multi View Representations and the Data Representation Techniques.

International audience; Hyperspectral images (HSIs) are of crucial importance in order to better understand features from a large number of spectral channels. Restricted by its inner imaging mechanism, the spatial resolution is often limited for HSIs. To alleviate this issue, in this work, we propose a simple and efficient architecture of deep convolutional neural networks to fuse a low-resolution HSI (LR-HSI) and a high-resolution multispectral image (HR-MSI), yielding a high-resolution HSI (HR-HSI). The network is designed to preserve both spatial and spectral information thanks to a new architecture based on: 1) the use of the LR-HSI at the HR-MSI’s scale to get an output with satisfied spectral preservation and 2) the application of the attention and pixelShuffle modules to extract information, aiming to output high-quality spatial details. Finally, a plain mean squared error loss function is used to measure the performance during the training. Extensive experiments demonstrate that the proposed network architecture achieves the best performance (both qualitatively and quantitatively) compared with recent state-of-the-art HSI super-resolution approaches. Moreover, other significant advantages can be pointed out by the use of the proposed approach, such as a better network generalization ability, a limited computational burden, and the robustness with respect to the number of training samples. Please find the source code and pretrained models from https://liangjiandeng.github.io/Projects_Res/HSRnet_2021tnnls.html .

Event cameras sense brightness changes in each pixel and yield asynchronous event streams instead of producing intensity images. They have distinct advantages over conventional cameras, such as a high dynamic range (HDR) and no motion blur. To take advantage of event cameras with existing image-based algorithms, a few methods have been proposed to reconstruct images from event streams. However, the output images have a low resolution (LR) and are unrealistic. Low-quality outputs stem from broader applications of event cameras, where high-quality and high-resolution (HR) images are needed. In this work, we consider the problem of reconstructing and super-resolving images from LR events when no ground truth (GT) HR images and degradation models are available. We propose a novel end-to-end joint framework for single image reconstruction and super-resolution from LR event data. Our method is primarily unsupervised to handle the absence of real inputs from GT and deploys adversarial learning. To train our framework, we constructed an open dataset, including simulated events and real-world images. The use of the dataset boosts the network performance, and the network architectures and various loss functions in each phase help improve the quality of the resulting image. Various experiments showed that our method surpasses the state-of-the-art LR image reconstruction methods for real-world and synthetic datasets. The experiments for super-resolution (SR) image reconstruction also substantiate the effectiveness of the proposed method. We further extended our method to more challenging problems of HDR, sharp image reconstruction, and color events. In addition, we demonstrate that the reconstruction and super-resolution results serve as intermediate representations of events for high-level tasks, such as semantic segmentation, object recognition, and detection. We further examined how events affect the outputs of the three phases and analyze our method's efficacy through an ablation study.

Vehicle Re-ID has recently attracted enthusiastic attention due to its potential applications in smart city and urban surveillance. However, it suffers from large intra-class variation caused by view variations and illumination changes, and inter-class similarity especially for different identities with the similar appearance. To handle these issues, in this paper, we propose a novel deep network architecture, which guided by meaningful attributes including camera views, vehicle types and colors for vehicle Re-ID. In particular, our network is end-to-end trained and contains three subnetworks of deep features embedded by the corresponding attributes (i.e., camera view, vehicle type and vehicle color). Moreover, to overcome the shortcomings of limited vehicle images of different views, we design a view-specified generative adversarial network to generate the multi-view vehicle images. For network training, we annotate the view labels on the VeRi-776 dataset. Note that one can directly adopt the pre-trained view (as well as type and color) subnetwork on the other datasets with only ID information, which demonstrates the generalization of our model. Extensive experiments on the benchmark datasets VeRi-776 and VehicleID suggest that the proposed approach achieves the promising performance and yields to a new state-of-the-art for vehicle Re-ID.

Cartoon is a common form of art in our daily life and automatic generation of cartoon images from photos is highly desirable. However, state-of-the-art single-style methods can only generate one style of cartoon images from photos and existing multi-style image style transfer methods still struggle to produce high-quality cartoon images due to their highly simplified and abstract nature. In this article, we propose a novel multi-style generative adversarial network (GAN) architecture, called MS-CartoonGAN, which can transform photos into multiple cartoon styles. MS-CartoonGAN uses only unpaired photos and cartoon images of multiple styles for training. To achieve this, we propose to use (1) a hierarchical semantic loss with sparse regularization to retain semantic content and recover flat shading in different abstract levels, (2) a new edge-promoting adversarial loss for producing fine edges, and (3) a style loss to enhance the difference between output cartoon styles and make training process more stable. We also develop a multi-domain architecture, where the generator consists of a shared encoder and multiple decoders for different cartoon styles, along with multiple discriminators for individual styles. By observing that cartoon images drawn by different artists have their unique styles while sharing some common characteristics, our shared network architecture exploits the common characteristics of cartoon styles, achieving better cartoonization and being more efficient than single-style cartoonization. We show that our multi-domain architecture can theoretically guarantee to output desired multiple cartoon styles. Through extensive experiments including a user study, we demonstrate the superiority of the proposed method, outperforming state-of-the-art single-style and multi-style image style transfer methods.

The subtleties of human perception, as measured by vision scientists through the use of psychophysics, are important clues to the internal workings of visual recognition. For instance, measured reaction time can indicate whether a visual stimulus is easy for a subject to recognize, or whether it is hard. In this paper, we consider how to incorporate psychophysical measurements of visual perception into the loss function of a deep neural network being trained for a recognition task, under the assumption that such information can enforce consistency with human behavior. As a case study to assess the viability of this approach, we look at the problem of handwritten document transcription. While good progress has been made towards automatically transcribing modern handwriting, significant challenges remain in transcribing historical documents. Here we describe a general enhancement strategy, underpinned by the new loss formulation, which can be applied to the training regime of any deep learning-based document transcription system. Through experimentation, reliable performance improvement is demonstrated for the standard IAM and RIMES datasets for three different network architectures. Further, we go on to show feasibility for our approach on a new dataset of digitized Latin manuscripts, originally produced by scribes in the Cloister of St. Gall in the the 9th century.

It is challenging to train deep spiking neural networks (SNNs) directly due to the difficulties associated with the non-differentiable neuron model. In this work, an end-to-end learning algorithm based on discrete current-based Leaky Integrate-and-Fire (C-LIF) neuron model and surrogate gradient is proposed to leverage the encoder-decoder network architecture to train deep SNNs directly. The proposed algorithm is capable of learning deep spatio-temporal features relying on current time step only, and several acceleration techniques including backward phase skipping and layer-wise FreezeOut are proposed to accelerate the training. Experimental results show that the proposed learning algorithm achieved the classification accuracies of 98.40% and 95.83% on the dynamic neuromorphic datasets MNIST-DVS and DVS-Gestures, respectively, and of 99.58% and 95.97% on static vision datasets MNIST and SVHN, respectively, which are comparable to the existing state-of-the-art results. The training speed was accelerated by up to 49.5% on MNIST and 36.6% on DVS-Gestures with the proposed acceleration techniques while maintaining the same level of accuracy.

The power distribution systems fault root cause classification is an important but challenging problem. Traditional classifiers fail to achieve high accuracy and good generalization performance due to data insufficiency. A large volume of unlabeled data is available, which can be utilized to improve classification performance. This paper proposes a novel classifier called Robust Semi-Supervised Prototypical Network (RSSPN) based on Prototypical Network architecture and semi-supervised learning to address this issue. The proposed method can mine information from unlabeled data to improve the generalization ability and classification accuracy. Furthermore, RSSPN adopts the idea of meta-learning to obtain the few-shot learning" ability for identifying new fault classes using very few samples encountered during the operation and update online. Experiments have been conducted on a dataset consisting of 1152 labeled samples belonging to 12 different classes and 10000 unlabeled samples. The accuracy of the proposed method is significantly better than the traditional classifiers.

In recent years, physiological signal based authentication has shown great promises,for its inherent robustness against forgery. Electrocardiogram (ECG) signal, being the most widely studied biosignal, has also received the highest level of attention in this regard. It has been proven with numerous studies that by analyzing ECG signals from different persons, it is possible to identify them, with acceptable accuracy. In this work, we present, EDITH, a deep learning-based framework for ECG biometrics authentication system. Moreover, we hypothesize and demonstrate that Siamese architectures can be used over typical distance metrics for improved performance. We have evaluated EDITH using 4 commonly used datasets and outperformed the prior works using less number of beats. EDITH performs competitively using just a single heartbeat (96-99.75% accuracy) and can be further enhanced by fusing multiple beats (100% accuracy from 3 to 6 beats). Furthermore, the proposed Siamese architecture manages to reduce the identity verification Equal Error Rate (EER) to 1.29%. A limited case study of EDITH with real-world experimental data also suggests its potential as a practical authentication system., Comment: Preprint

Unknown surface material classification can inform a robot about material properties, enabling it to interact with environments appropriately. Recent research has leveraged multimodal data using deep learning to improve the performance of surface material classification. In this paper, we present a deep learning model, multimodal temporal convolutional neural network (MTCNN), which integrates energy spectrum, dilated convolutions, and sequence poolings into a unified network architecture. The proposed model can learn material representations from auditory and multi-tactile (i.e., acceleration, normal force, and friction force) data generated by dragging a tool along surfaces, and distinguish unknown object surface materials into categories. For surface material data collection, a tool is also designed to detect different object surfaces. The performance of MTCNN is evaluated on a public dataset and the highest classification accuracy is 87.55%. A robotic curling example is provided to illustrate how the presented model helps the robot in manipulation.

Screening mammograms is the gold standard for detecting breast cancer early. While a good amount of work has been performed on mammography image classification, especially with deep neural networks, there has not been much exploration into the confidence or uncertainty measurement of the classification. In this paper, we propose a confidence measure-based evaluation metric for breast cancer screening. We propose a modular network architecture, where a traditional neural network is used as a feature extractor with transfer learning, followed by a simple Bayesian neural network. Utilizing a two-stage approach helps reducing the computational complexity, making the proposed framework attractive for wider deployment. We show that by providing the medical practitioners with a tool to tune two hyperparameters of the Bayesian neural network, namely, fraction of sampled number of networks and minimum probability, the framework can be adapted as needed by the domain expert. Finally, we argue that instead of just a single number such as accuracy, a tuple (accuracy, coverage, sampled number of networks, and minimum probability) can be utilized as an evaluation metric of our framework. We provide experimental results on the CBIS-DDSM dataset, where we show the trends in accuracy-coverage tradeoff while tuning the two hyperparameters. We also show that our confidence tuning results in increased accuracy with a reduced set of images with high confidence when compared to the baseline transfer learning. To make the proposed framework readily deployable, we provide (anonymized) source code with reproducible results at https://git.io/JvRqE., Comment: To be presented at the IEEE ICHI 2020

Countries have different specifications for License Plates (LPs), therefore developing one Automatic license plate recognition (ALPR) system that works well for all LPs types is a difficult task. This paper aims to develop an accurate ALPR for Jordanian LPs. Two-stage Convolutional Neural Networks (CNNs) are used in the proposed approach, the CNNs are based on the YOLO3 framework. The sizes of LPs' characters are very small compared with the frame size, therefore the YOLO3 network architecture is modified to a shallow network to detect small objects. The proposed approach uses temporal information from different frames to remove false predictions. A set of arrays data structure is used to track the vehicles’ LPs and eliminate incorrect ones. To my knowledge, the proposed approach represents the first end-to-end Jordanian ALPR that processes video stream in real-time. To my knowledge, there is no dataset for Jordanian license plates, therefore this paper proposes a new dataset called JALPR dataset. The dataset is available online and includes many real videos for moving vehicles in Jordan. Two well-known commercial software packages are used for comparisons. The experimental results in real videos from YouTube show that the proposed approach is very efficient in recognizing the Jordanian license plates and achieved 87% recognition accuracy, whereas the commercial systems have recognition accuracies that are less than 81%.

This study is concerned with the concept of Ensemble Fuzzy Radial Basis Function Neural Networks (EFRBFNN) endowed with clustering techniques, polynomial-based LSE/WLSE learning as well as their optimization implemented by Multi-Objective Particle Swarm Optimization (MOPSO). The network architecture of the proposed EFRBFNN classifier is considered three types of clustering techniques and four forms of regression polynomials. The proposed classifier can not only fully capture the local distribution (feature) information contained in the data, but also choose effective network architecture for improving classification performance and compactness. Three classes of clustering techniques (such as iterative self-organizing data analysis techniques algorithm (ISODATA), affinity propagation (AP), and extended K-means) are considered to effectively produce membership degrees based on feature information extraction suitable to fit data characteristics. Least Square Error Estimation (LSE) and Weighted Least Square Error Estimation (WLSE)-based learning with four types of regression polynomials are considered to estimate to coefficients of polynomials. MOPSO is exploited for choosing the effective architecture among different clustering techniques, polynomials, and polynomial-based learning with multi-objective parametric optimization. MOPSO helps achieve a sound compromise between the preferred classification performance and the compactness realized with the aid of four objective functions such as a) classification accuracy (CA), b) complexity of clustering, c) complexity of polynomial and d) sum of squared coefficients (SSC). The improvement and effectiveness of the proposed network architecture are quantified on a basis of the comprehensive experimental results and also a comparative analysis is offered to demonstrate the superiority of the proposed classifier.

As an essential problem in computer vision, salient object detection (SOD) has attracted an increasing amount of research attention over the years. Recent advances in SOD are predominantly led by deep learning-based solutions (named deep SOD). To enable in-depth understanding of deep SOD, in this paper, we provide a comprehensive survey covering various aspects, ranging from algorithm taxonomy to unsolved issues. In particular, we first review deep SOD algorithms from different perspectives, including network architecture, level of supervision, learning paradigm, and object-/instance-level detection. Following that, we summarize and analyze existing SOD datasets and evaluation metrics. Then, we benchmark a large group of representative SOD models, and provide detailed analyses of the comparison results. Moreover, we study the performance of SOD algorithms under different attribute settings, which has not been thoroughly explored previously, by constructing a novel SOD dataset with rich attribute annotations covering various salient object types, challenging factors, and scene categories. We further analyze, for the first time in the field, the robustness of SOD models to random input perturbations and adversarial attacks. We also look into the generalization and difficulty of existing SOD datasets. Finally, we discuss several open issues of SOD and outline future research directions., Published on IEEE TPAMI. All the saliency prediction maps, our constructed dataset with annotations, and codes for evaluation are publicly available at \url{https://github.com/wenguanwang/SODsurvey}

Wireless Capsule Endoscopy (WCE) has been widely accepted due to its painless method of imaging the entire gastrointestinal tract. In this paper, we propose deep Convolutional Neural Network(CNN) for automatic discrimination of ulcers on different ratios of augmented datasets ranging from 1000 to 10000 WCE images comprising of ulcer and non-ulcer images. A detailed investigation of network configuration for various nodes and depth were performed. The proposed network architecture of four convolutional layers with ( 3 * 3 ) convolutional filters demonstrated significant improvement in terms of performance. The WCE images were obtained from publicly available WCE datasets and real-time WCE video frames. The test results were subjected to hyper-parameter optimization for various tweaking parameters such as epochs, pooling schemes, learning rate, number of layers, optimizer, activation functions and drop out scheme. The experimental results were compared with ten different machine learning classifiers, demonstrating higher prediction performance.

The increasing throughput of experiments in biomaterials research makes automatic techniques more and more necessary. Among all the characterization methods, microscopy makes fundamental contributions to biomaterials science where precisely focused images are the basis of related research. Although automatic focusing has been widely applied in all kinds of microscopes, defocused images can still be acquired now and then due to factors including background noises of materials and mechanical errors. Herein, we present a deep-learning-based method for the automatic sorting and reconstruction of defocused cell images. First, the defocusing problem is illustrated on a high-throughput cell microarray. Then, a comprehensive dataset of phase-contrast images captured from varied conditions containing multiple cell types, magnifications, and substrate materials is prepared to establish and test our method. We obtain high accuracy of over 0.993 on the dataset using a simple network architecture that requires less than half of the training time compared with the classical ResNetV2 architecture. Moreover, the subcellular-level reconstruction of heavily defocused cell images is achieved with another architecture. The applicability of the established workflow in practice is finally demonstrated on the high-throughput cell microarray. The intelligent workflow does not require a priori knowledge of focusing algorithms, possessing widespread application value in cell experiments concerning high-throughput or time-lapse imaging., Graphical abstract Image 1, Highlights • An automatic workflow for the sorting and reconstruction of defocused images. • Representative data containing varied cells, substrate materials, and magnifications. • Effectiveness of the workflow in high-throughput experiments. • Publicly accessible image datasets and codes.

This paper explores the use of ambient radio frequency (RF) signals for human presence detection through deep learning. Using WiFi signal as an example, we demonstrate that the channel state information (CSI) obtained at the receiver contains rich information about the propagation environment. Through judicious pre-processing of the estimated CSI followed by deep learning, reliable presence detection can be achieved. Several challenges in passive RF sensing are addressed. With presence detection, how to collect training data with human presence can have a significant impact on the performance. This is in contrast to activity detection when a specific motion pattern is of interest. A second challenge is that RF signals are complex-valued. Handling complex-valued input in deep learning requires careful data representation and network architecture design. Finally, human presence affects CSI variation along multiple dimensions; such variation, however, is often masked by system impediments such as timing or frequency offset. Addressing these challenges, the proposed learning system uses pre-processing to preserve human motion induced channel variation while insulating against other impairments. A convolutional neural network (CNN) properly trained with both magnitude and phase information is then designed to achieve reliable presence detection. Extensive experiments are conducted. Using off-the-shelf WiFi devices, the proposed deep learning based RF sensing achieves near perfect presence detection during multiple extended periods of test and exhibits superior performance compared with leading edge passive infrared sensors. Comparison with existing RF based human presence detection also demonstrates its robustness in performance, especially when deployed in a completely new environment., Source code and datasets are available at Github: https://github.com/bigtreeyanger/presence_detection_cnn

Neural networks have become state-of-the-art computer vision tools for tasks that learn implicit representations of geometrical scenes. This paper proposes a two-part network architecture that expl...

This paper introduces BReG-NeXt, a residual-based network architecture using a function wtih bounded derivative instead of a simple shortcut path (a.k.a. identity mapping) in the residual units for automatic recognition of facial expressions based on the categorical and dimensional models of affect. Compared to ResNet, our proposed adaptive complex mapping results in a shallower network with less numbers of training parameters and floating point operations per second (FLOPs). Adding trainable parameters to the bypass function further improves fitting and training the network and hence recognizing subtle facial expressions such as contempt with a higher accuracy. We conducted comprehensive experiments on the categorical and dimensional models of affect on the challenging in-the-wild databases of AffectNet, FER2013, and Affect-in-Wild. Our experimental results show that our adaptive complex mapping approach outperforms the original ResNet consisting of a simple identity mapping as well as other state-of-the-art methods for Facial Expression Recognition (FER). Various metrics are reported in both affect models to provide a comprehensive evaluation of our method. In the categorical model, BReG-NeXt-50 with only 3.1M training parameters and 15 MFLOPs, achieves 68.50% and 71.53% accuracy on AffectNet and FER2013 databases, respectively. In the dimensional model, BReG-NeXt achieves 0.2577 and 0.2882 RMSE value on AffectNet and Affect-in-Wild databases, respectively., To appear in IEEE Transactions on Affective Computing journal

Facial Attribute Classification (FAC) has attracted increasing attention in computer vision and pattern recognition. However, state-of-the-art FAC methods perform face detection/alignment and FAC independently. The inherent dependencies between these tasks are not fully exploited. In addition, most methods predict all facial attributes using the same CNN network architecture, which ignores the different learning complexities of facial attributes. To address the above problems, we propose a novel deep multi-task multi-label CNN, termed DMM-CNN, for effective FAC. Specifically, DMM-CNN jointly optimizes two closely-related tasks (i.e., facial landmark detection and FAC) to improve the performance of FAC by taking advantage of multi-task learning. To deal with the diverse learning complexities of facial attributes, we divide the attributes into two groups: objective attributes and subjective attributes. Two different network architectures are respectively designed to extract features for two groups of attributes, and a novel dynamic weighting scheme is proposed to automatically assign the loss weight to each facial attribute during training. Furthermore, an adaptive thresholding strategy is developed to effectively alleviate the problem of class imbalance for multi-label learning. Experimental results on the challenging CelebA and LFWA datasets show the superiority of the proposed DMM-CNN method compared with several state-of-the-art FAC methods.

Prevailing deep methods for image recognition require massive labeled samples in each visual category for training. However, large amounts of data annotations are time-consuming, and some uncommon categories only have rare samples available. For this issue, we focus on more challenging few-shot learning (FSL) task, where just few labeled images are used in the training stage. Existing FSL models are constructed with various convolutional neural networks (CNNs), which are trained on an auxiliary base dataset and evaluated for new few-shot predictions on a novel dataset. The performance of these models is difficult to break through because of the domain shift between base and novel datasets and the monotonous network architectures. Considering that, we propose a novel automatic attribute consistent network called Auto-ACNet to overcome the above problems. On one hand, Auto-ACNet utilizes the attribute information about base and novel categories to guide the procedure of representation learning. It introduces the consistent and non-consistent subnets to capture the common and different attributes of image pair, which helps to mitigate the domain shift problem. On the other hand, the architecture of Auto-ACNet is searched with the popular neural architecture search (NAS) technique DARTS, for obtaining a superior FSL network automatically. And the DARTS’s search space is improved by adding the position-aware module to extract the attribute characteristics better. Extensive experimental results on two datasets indicate that the proposed Auto-ACNet achieves significant improvement over the state-of-the-art competitors in this literature.