Image Classification Algorithm Research Articles

An approach for identifying historic village using deep learning

This paper aims to propose an approach to automatically identify historic villages from remote sensing images based on deep learning algorithm and accurately calculate the villages’ geographical coordinates. Experimental datasets of Conghua, a typical region in fast development that retains many historic villages, are designated for training and testing. Comparison experiments of two recognition models, image classification and object detection, are designed to obtain the most suitable identification algorithm. GIS platform is adopted to visualize the distribution of the historic villages. The results show that first, the recognition accuracy of the image classification algorithm is 90.79%. However, visualization of test results shows the identified area is not a village but a surrounding. Second, the recognition accuracy of an object detection algorithm can reach 95.61%, which indicates that the algorithm is accurate and efficient. Third, by using the Historical-Modern tag as a filter, a village with a certain proportion of historic features according to specific requirements may be discriminated. Finally, 1531 historic villages in Conghua area were identified by the preferred algorithm, and their spatial locations were marked. This research will extend the detection of remote sensing image targets of deep learning algorithms from single buildings to group patterns and complex ground objects, so as to promote the integration of heritage conservation and artificial intelligence research. This time-efficiency approach can provide strong support for the discovery and field investigation of historic villages facing fast development and provide a scientific basis for the formulation of conservation policies.Article HighlightsDeep learning is applied to the protection of the cultural heritage of historic villages.Comparative experiments of different algorithms are designed to analyse their applicability in historic village recognition. A recognition rate of up to 95.61% is achieved.The visualization of recognition results is important for understanding the relationship between historic villages and nature, and historic village conservation.

A Deep Learning Approach for Detecting Stroke from Brain CT Images Using OzNet.

A brain stroke is a life-threatening medical disorder caused by the inadequate blood supply to the brain. After the stroke, the damaged area of the brain will not operate normally. As a result, early detection is crucial for more effective therapy. Computed tomography (CT) images supply a rapid diagnosis of brain stroke. However, while doctors are analyzing each brain CT image, time is running fast. This circumstance may lead to result in a delay in treatment and making errors. Therefore, we targeted the utilization of an efficient artificial intelligence algorithm in stroke detection. In this paper, we designed hybrid algorithms that include a new convolution neural networks (CNN) architecture called OzNet and various machine learning algorithms for binary classification of real brain stroke CT images. When we classified the dataset with OzNet, we acquired successful performance. However, for this target, we combined it with a minimum Redundancy Maximum Relevance (mRMR) method and Decision Tree (DT), k-Nearest Neighbors (kNN), Linear Discriminant Analysis (LDA), Naïve Bayes (NB), and Support Vector Machines (SVM). In addition, 4096 significant features were obtained from the fully connected layer of OzNet, and we reduced the dimension of features from 4096 to 250 using the mRMR method. Finally, we utilized these machine learning algorithms to classify important features. As a result, OzNet-mRMR-NB was an excellent hybrid algorithm and achieved an accuracy of 98.42% and AUC of 0.99 to detect stroke from brain CT images.

Hyper-sausage coverage function neuron model and learning algorithm for image classification

Recently, deep neural networks (DNNs) promote mainly by network architectures and loss functions; however, the development of neuron models has been quite limited. In this study, inspired by the mechanism of human cognition, a hyper-sausage coverage function (HSCF) neuron model possessing a high flexible plasticity. Then, a novel cross-entropy and volume-coverage (CE_VC) loss is defined, which compresses the volume of the hyper-sausage to the hilt, and helps alleviate confusion among different classes, thus ensuring the intra-class compactness of the samples. Finally, a divisive iteration method is introduced, which considers each neuron model as a weak classifier, and iteratively increases the number of weak classifiers. Thus, the optimal number of the HSCF neuron is adaptively determined and an end-to-end learning framework is constructed. In particular, to improve the classification performance, the HSCF neuron can be applied to classical DNNs. Comprehensive experiments on eight datasets in several domains demonstrate the effectiveness of the proposed method. The proposed method exhibits the feasibility of boosting DNNs with neuron plasticity and provides a novel perspective for further developments in DNNs. The source code is available at https://github.com/Tough2011/HSCFNet.git .

Feature Extraction using Histogram of Oriented Gradients for Image Classification in Maize Leaf Diseases

The paper presents feature extraction methods and classification algorithms used to classify maize leaf disease images. From maize disease images, features are extracted and passed to the machine learning classification algorithm to identify the possible disease based on the features detected using the feature extraction method. The maize disease images used include images of common rust, leaf spot, and northern leaf blight and healthy images. An evaluation was done for the feature extraction method to see which feature extraction method performs best with image classification algorithms. Based on the evaluation, the outcomes revealed Histogram of Oriented Gradients performed best with classifiers compared to KAZE and Oriented FAST and rotated BRIEF. The random forest classifier emerged the best in terms of image classification, based on four performance metrics which are accuracy, precision, recall, and F1-score. The experimental outcome indicated that the random forest had 0.74 accuracy, 0.77 precision, 0.77 recall, and 0.75 F1-score.

Research on improved intestinal image classification for LARS based on ResNet.

Low anterior rectal resection is an effective way to treat rectal cancer at present, but it is easy to cause low anterior resection syndrome after surgery; so, a comprehensive diagnosis of defecation and pelvic floor function must be carried out. There are few studies on the classification of diagnoses in the field of intestinal diseases. In response to these outstanding problems, this research will focus on the design of the intestinal function diagnosis system and the image processing and classification algorithm of the intestinal wall to verify an efficient fusion method, which can be used to diagnose the intestinal diseases in clinical medicine. The diagnostic system designed in this paper makes up for the singleness of clinical monitoring methods. At the same time, the Res-SVDNet neural network model is used to solve the problems of small intestinal image samples and network overfitting, and achieve efficient fusion diagnosis of intestinal diseases in patients. Different models were used to compare experiments on the constructed datasets to verify the applicability of the Res-SVDNet model in intestinal image classification. The accuracy of the model was 99.54%, which is several percentage points higher than other algorithm models.

Defects recognition of strain clamps based on attentional enhancement mechanism

The crimping quality of tension clamp in overhead transmission line is directly related to the safe and stable operation of transmission line. existing detection methods are low efficiency or availability and developing a timely and effective defect detecting method for strain clamp is a top priority. An X-ray image classification algorithm based on the attentional enhancement mechanism is proposed to identify the defects of different types of strain clamps. DenseNet structure is used as the backbone network to multiplex the X-ray image features, SENet channel is used to enhance the attention to the defect site, and PSANet is used to replace the 3×3 convolution layer in DenseBlock in DenseNet to highlight the features of defects at different scales. Research shows that the average recognition accuracy of the proposed algorithm is 88.1%, which is better than the existing methods.

Underwater Image Classification Algorithm Based on Convolutional Neural Network and Optimized Extreme Learning Machine

In order to deal with the target recognition in the complex underwater environment, we carried out experimental research. This includes filtering noise in the feature extraction stage of underwater images rich in noise, or with complex backgrounds, and improving the accuracy of target classification in the recognition process. This paper discusses our contribution to improving the accuracy of underwater target classification. This paper proposes an underwater target classification algorithm based on the improved flow direction algorithm (FDA) and search agent strategy, which can simultaneously optimize the weight parameters, bias parameters, and super parameters of the extreme learning machine (ELM). As a new underwater target classifier, it replaces the full connection layer in the traditional classification network to build a classification network. In the first stage of the network, the DenseNet201 network pre-trained by ImageNet is used to extract features and reduce dimensions of underwater images. In the second stage, the optimized ELM classifier is trained and predicted. In order to weaken the uncertainty caused by the random input weight and offset of the introduced ELM, the fuzzy logic, chaos initialization, and multi population strategy-based flow direction algorithm (FCMFDA) is used to adjust the input weight and offset of the ELM and optimize the super parameters with the search agent strategy at the same time. We tested and verified the FCMFDA-ELM classifier on Fish4Knowledge and underwater robot professional competition 2018 (URPC 2018) datasets, and achieved 99.4% and 97.5% accuracy, respectively. The experimental analysis shows that the FCMFDA-ELM underwater image classifier proposed in this paper has a greater improvement in classification accuracy, stronger stability, and faster convergence. Finally, it can be embedded in the recognition process of underwater targets to improve the recognition performance and efficiency.

Convolutional neural network based on fast Fourier transform and gramian angle field for fault identification of HVDC transmission line

Flexible DC grid puts forward high-speed dynamic, and high-reliability requirements for DC transmission line protection. How to identify DC line faults ultra-fast and reliably is one of the critical technologies for the further development of the flexible DC grid. To solve this problem, a new DC transmission line fault identification scheme based on the convolutional neural network (CNN) based on fast Fourier transform (FFT) and gramian angular field (GAF) is proposed. The proposed scheme aims to further reduce the identification time of DC transmission line fault types under the requirements of rapidity and accuracy of DC transmission line protection. In this scheme, the obtained feature data are processed by FFT and GAF to obtain the feature images of different fault types corresponding to DC transmission lines. Then the CNN is used to classify the feature images to realize the fault identification of DC transmission lines. Compared with previous methods, this method can apply the mature image classification algorithm in artificial intelligence algorithm to DC transmission line fault identification and realize the rapid identification of DC transmission line fault types. The simulation results show that this method can only use the fault data within 1.5 ms after the fault to achieve 100% accuracy of fault recognition. When the fault data acquisition window is reduced to 1 ms it still has 99.88% classification accuracy, which has certain advantages in the speed and accuracy of DC transmission line fault recognition.

An Image Classification and Retrieval Algorithm for Product Display in E-Commerce Transactions

E-commerce has unmatched advantages over conventional ways of consumption, but it also restricts the amount of contact that customers have with the things they are purchasing. The first step in the transaction product information flow is an effective transaction product display image, which is crucial for product sales. The influence of the emotions present in the current transaction product display image on the buyer's purchase intention has not been established, and as the number of commodity classes rises, there is also a lack of scientific guidance regarding the buyer's efficient retrieval of transaction products. Therefore, this work investigates image classification and retrieval methods for product presentation in e-commerce transactions. A robust network architecture was created for the e-commerce transaction product display and image classification. The image polarity emotion feature extraction backbone module, the polarity emotion intensity perception module, and the emotion feature fusion classification module are the three specific components of the model. The network training approach is an innovation to address the issue that the polarity emotion intensity cannot be effectively conveyed in the e-commerce transaction product display image. The research improves the design of the similarity retrieval algorithm for e-commerce transaction product display images, which increased the retrieval effectiveness for buyers. The correctness of the classification model and retrieval approach was confirmed by the experimental results.

A Survey of Image Classification Algorithms based on Convolution Neural Network

With the deep learning (DL) sweeping the world. Traditional image classification methods are difficult to process huge image data and cannot meet people's requirements for image classification accuracy and speed. The image classification method based on convolutional neural network (CNN) breaks through the bottle neck of traditional image classification methods and becomes the mainstream algorithm of image classification at present, how to effectively use convolutional neural network to classify images has become a hot research topic in the field of computer vision at home and abroad. Convolutional neural network (CNN) has performed well in image classification and segmentation, target detection and other applications, and its powerful feature learning and feature expression capabilities are increasingly respected by researchers. However, CNN still has a few problems, such as incomplete feature extraction and overfitting of sample training. In view of these problems, after in-depth research on the application of convolutional neural network in image processing, this paper gives the mainstream structure model, advantages and disadvantages, time/space complexity, problems that may be encountered in the model training process and corresponding solutions used in image classification based on convolutional neural network. Through the overview of the research status of CNN model in image classification, it provides suggestions for the further development and research direction of CNN.

EHDFL: Evolutionary hybrid domain feature learning based on windowed fast Fourier convolution pyramid for medical image classification

With the development of modern medical technology, medical image classification has played an important role in medical diagnosis and clinical practice. Medical image classification algorithms based on deep learning emerge in endlessly, and have achieved amazing results. However, most of these methods ignore the feature representation based on frequency domain, and only focus on spatial features. To solve this problem, we propose a hybrid domain feature learning (HDFL) module based on windowed fast Fourier convolution pyramid, which combines the global features with a wide range of receptive fields in frequency domain and the local features with multiple scales in spatial domain. In order to prevent frequency leakage, we construct a Windowed Fast Fourier Convolution (WFFC) structure based on Fast Fourier Convolution (FFC). In order to learn hybrid domain features, we combine ResNet, FPN, and attention mechanism to construct a hybrid domain feature learning module. In addition, a super-parametric optimization algorithm is constructed based on genetic algorithm for our classification model, so as to realize the automation of our super-parametric optimization. We evaluated the newly published medical image classification dataset MedMNIST, and the experimental results show that our method can effectively learning the hybrid domain feature information of frequency domain and spatial domain.

Pneumonia Detection Using Novel Deep Learning Techniques

Pneumonia is one of the leading infectious disease that can kill children and elderly people around the world. The development of an automated system to detect pneumonia would be advantageous, especially to enable treatment of this disease in remote areas without much delay. pneumonia is an lung infectious disease which mainly affects the small air sacs known as alveoli. The main symptoms of pneumonia includes cough, fever and breathing problems . Aged people, children and persons who have medical problems are the main victims of this disease. Around 450 million people are affected by this disease on an average of each and every year. The most commonly used technique for diagnosing this disease is chest X-ray imaging. Chest X-ray examination is complex procedure to detect the disease because it involves lots of vulnerabilities. With latest advances in technologies we can use deep learning algorithms to detect the disease using chest x ray images. To deal with the scarcity of data, we used Deep Transfer Learning and designed the Hybrid Algorithms. The images of the chest X-rays were fed into the individual algorithms for training purposes. Parallel Deep Feature Extractors are used in conjunction with various algorithms. For classifying chest X-ray images into normal and pneumonia, Here we are proposing an hybrid model based on VGG16, VGG19, CNN, and Mobile Net networks. Individual image classification algorithms were combined to form a hybrid model .In comparison to individual algorithms, the new Hybrid Model with Deep Learning achieved higher accuracy than existing methods.

Researches Advanced in Deep Learning based Image Classification

Image recognition has always been a popular research topic in computer vision, whose basic task is to learn a model to predict the category of a given image. Early image classification algorithms mainly relied on handcrafted features, while their classification results often failed to meet practical application requirements due to the limitation of handcrafted features expressiveness ability. Thanks to the rapid development of deep learning, image recognition algorithms based on convolutional neural networks have achieved great success. Generally, stacking network layers can improve the prediction accuracy, while increasing the network depth can also lead to problems such as gradient disappearance, gradient explosion, and degradation. In recent years, due to its powerful representation ability, Transformer-based image classification algorithms have achieved new breakthroughs in recognition accuracy. This paper first introduces the classic deep learning algorithms in the field of image classification, including networks such as AlexNet, GoogLeNet, VGG, and ResNet. Meanwhile, the visual transformer (ViT) and the data-efficient image transformer are further introduced to handle image classification tasks. In addition, this paper analyzes the application and development of these two models in image classification, classifies the different models, and analyzes their advantages and disadvantages.

Performance and Accuracy Comparisons of Classification Methods and Perspective Solutions for UAV-Based Near-Real-Time "Out of the Lab" Data Processing.

Today, integration into automated systems has become a priority in the development of remote sensing sensors carried on drones. For this purpose, the primary task is to achieve real-time data processing. Increasing sensor resolution, fast data capture and the simultaneous use of multiple sensors is one direction of development. However, this poses challenges on the data processing side due to the increasing amount of data. Our study intends to investigate how the running time and accuracy of commonly used image classification algorithms evolve using Altum Micasense multispectral and thermal acquisition data with GSD = 2 cm spatial resolution. The running times were examined for two PC configurations, with a 4 GB and 8 GB DRAM capacity, respectively, as these parameters are closer to the memory of NRT microcomputers and laptops, which can be applied "out of the lab". During the accuracy assessment, we compared the accuracy %, the Kappa index value and the area ratio of correct pixels. According to our results, in the case of plant cover, the Spectral Angles Mapper (SAM) method achieved the best accuracy among the validated classification solutions. In contrast, the Minimum Distance (MD) method achieved the best accuracy on water surface. In terms of temporality, the best results were obtained with the individually constructed decision tree classification. Thus, it is worth developing these two directions into real-time data processing solutions.

An enhanced multiscale generation and depth-perceptual loss-based super-resolution network for prostate ultrasound images

Computer-aided diagnosis of prostate ultrasound images is gradually being combined with deep learning to help detect and treat diseases. However, ultrasound images of the prostate have problems such as low resolution and unbalanced categories. In addition, the current image classification algorithms have difficulty with classification network performance due to insufficient data volume. To solve these problems, inspired by CycleGAN, we propose an enhanced multiscale generation and depth-perceptual loss-based super-resolution (SR) network for prostate ultrasound images (EGDL-CycleGAN). We study and improve the generative network and perceptual loss of CycleGAN. In this work we achieve multiscale feature extraction through an improved generator, and utilize full-scale skip connections between encoder and decoder to capture fine-grained details and coarse-grained semantics at full scale. This effectively improves the performance of the generative network and makes the reconstruction effect better. We also use the residual structure for deep extraction of features to obtain perceptual loss, and add this to the network loss function for training the model. This enables the model to learn the global and local differences between the real and generated images. This approach pays more attention to the edge information and spatial information of the image, and provides relevant spatial information feedback to the generator to improve the generator’s ability to perceive consistent super-resolution. The method can enhance the prostate ultrasound image dataset and provide rich images for the next step in intelligence-assisted classification and diagnosis of prostate cancer ultrasound images. The evaluation of peak signal-to-noise ratio/structural similarity and visual effects against the benchmark of our datasets illustrates that our proposed approach is effective and superior to the bicubic classic image SR reconstruction algorithm, the SRGAN perception-driven method and the CycleGAN method applied to ultrasound images. In addition, the method of using the original dataset combined with the SR reconstruction image dataset can effectively improve the accuracy of the classification network in intelligence-assisted classification diagnosis of prostate cancer ultrasound images. In EfficientNetV2 the accuracy is improved from 0.843 to 0.867 and in Swin Transformer the accuracy is improved from 0.893 to 0.917.

Context-Aware Querying, Geolocalization, and Rephotography of Historical Newspaper Images

Newspapers contain a wealth of historical information in the form of articles and illustrations. Libraries and cultural heritage institutions have been digitizing their collections for decades to enable web-based access to and retrieval of information. A number of challenges arise when dealing with digitized collections, such as those of KBR, the Royal Library of Brussels (used in this study), which contain only page-level metadata, making it difficult to extract information from specific contexts. A context-aware search relies heavily on metadata enhancement. Therefore, when using metadata at the page level, it is even more challenging to geolocalize less-known landmarks. To overcome this challenge, we have developed a pipeline for geolocalization and visualization of historical photographs. The first step of this pipeline consists of converting page-level metadata to article-level metadata. In the next step, all articles with building images were classified based on image classification algorithms. Moreover, to correctly geolocalize historical photographs, we propose a hybrid approach that uses both textual metadata and image features. We conclude this research paper by addressing the challenge of visualizing historical content in a way that adds value to humanities research. It is noteworthy that a number of historical urban scenes are visualized using rephotography, which is notoriously challenging to get right. This study serves as an important step towards enriching historical metadata and facilitating cross-collection linkages, geolocalization, and the visualization of historical newspaper images. Furthermore, the proposed methodology is generic and can be used to process untagged photographs from social media, including Flickr and Instagram.

EP005 - An image classification algorithm for prostate cancer diagnosis on multiparametric transrectal ultrasound – Preliminary results of a prospective multicenter diagnostic accuracy trial

EP005 - An image classification algorithm for prostate cancer diagnosis on multiparametric transrectal ultrasound – Preliminary results of a prospective multicenter diagnostic accuracy trial

Balanced and Accurate Pseudo-Labels for Semi-Supervised Image Classification

Image classification by semi-supervised learning has recently become a hot spot, and the Co-Training framework is an important method of semi-supervised image classification. In the traditional Co-Training structure, the sub-networks will generate pseudo-labels for each other, and these pseudo-labels will further be used as a supervisory signal for model training. However, the pseudo-labels will hurt classification performance because of their low accuracy and unbalanced distribution. In this article, we are trying to solve the preceding two problems by designing the Balanced Module (BM) and Gaussian Mixture Module (GMM), and propose BAPS (the B alanced and A ccurate P seudo-labels for S emi-supervised image classification). In BM, the two sub-networks jointly predict the unlabeled images, then select the pseudo-labels with a high-confidence threshold to perform the balancing operation to obtain the initial samples with balanced distribution of each category. In GMM, referring to the common practice of the Learning from Noise Labels task, we use GMM to fit the loss distribution of images with pseudo-labels output by BM, then clean samples and noise samples are divided based on the observation that the loss of correctly labeled images is generally smaller than that of wrongly labeled ones. Through BM and GMM, pseudo-labels with balanced distribution and high accuracy are obtained for the subsequent model training process. Our model has achieved better classification accuracy than most state-of-the-art semi-supervised image classification algorithms on the CIFAR-10/100 and SVHN datasets, and further ablation experiments demonstrate the effectiveness of our BAPS. The source code of BAPS will be available at https://github.com/zhaojianaaa .

Geospatial-based machine learning techniques for land use and land cover mapping using a high-resolution unmanned aerial vehicle image

Bangladesh is primarily an agricultural country where technological advancement in the agricultural sector can ensure the acceleration of economic growth and ensure long-term food security. This research was conducted in the south-western coastal zone of Bangladesh, where rice is the main crop and other crops are also grown. Land use and land cover (LULC) classification using remote sensing techniques such as the use of satellite or unmanned aerial vehicle (UAV) images can forecast the crop yield and can also provide information on weeds, nutrient deficiencies, diseases, etc. to monitor and treat the crops. Depending on the reflectance received by sensors, remotely sensed images store a digital number (DN) for each pixel. Traditionally, these pixel values have been used to separate clusters and classify various objects. However, it frequently generates a lot of discontinuity in a particular land cover, resulting in small objects within a land cover that provide poor image classification output. It is called the salt-and-pepper effect. In order to classify land cover based on texture, shape, and neighbors, Pixel-Based Image Analysis (PBIA) and Object-Based Image Analysis (OBIA) methods use digital image classification algorithms like Maximum Likelihood (ML), K-Nearest Neighbors (KNN), k-means clustering algorithm, etc. to smooth this discontinuity. The authors evaluated the accuracy of both the PBIA and OBIA approaches by classifying the land cover of an agricultural field, taking into consideration the development of UAV technology and enhanced image resolution. For classifying multispectral UAV images, we used the KNN machine learning algorithm for object-based supervised image classification and Maximum Likelihood (ML) classification (parametric) for pixel-based supervised image classification. Whereas, for unsupervised classification using pixels, we used the K-means clustering technique. For image analysis, Near-infrared (NIR), Red (R), Green (G), and Blue (B) bands of a high-resolution ground sampling distance (GSD) 0.0125m UAV image was used in this research work. The study found that OBIA was 21% more accurate than PBIA, indicating 94.9% overall accuracy. In terms of Kappa statistics, OBIA was 27% more accurate than PBIA, indicating Kappa statistics accuracy of 93.4%. It indicates that OBIA provides better classification performance when compared to PBIA for the classification of high-resolution UAV images. This study found that by suggesting OBIA for more accurate identification of types of crops and land cover, which will help crop management, agricultural monitoring, and crop yield forecasting be more effective.

IMAGE CLASSIFIER RESILIENT TO ADVERSARIAL ATTACKS, FAULT INJECTIONS AND CONCEPT DRIFT – MODEL ARCHITECTURE AND TRAINING ALGORITHM

Context. The problem of image classification algorithms vulnerability to destructive perturbations has not yet been definitively resolved and is quite relevant for safety-critical applications. Therefore, object of research is the process of training and inference for image classifier that functioning under influences of destructive perturbations. The subjects of the research are model architecture and training algorithm of image classifier that provide resilience to adversarial attacks, fault injection attacks and concept drift.
 Objective. Stated research goal is to develop effective model architecture and training algorithm that provide resilience to adversarial attacks, fault injections and concept drift.
 Method. New training algorithm which combines self-knowledge distillation, information measure maximization, class distribution compactness and interclass gap maximization, data compression based on discretization of feature representation and semi-supervised learning based on consistency regularization is proposed.
 Results. The model architecture and training algorithm of image classifier were developed. The obtained classifier was tested on the Cifar10 dataset to evaluate its resilience over an interval of 200 mini-batches with a training and test size of mini-batch equals to 128 examples for such perturbations: adversarial black-box L∞-attacks with perturbation levels equal to 1, 3, 5 and 10; inversion of one randomly selected bit in a tensor for 10%, 30%, 50% and 60% randomly selected tensors; addition of one new class; real concept drift between a pair of classes. The effect of the feature space dimensionality on the value of the information criterion of the model performance without perturbations and the value of the integral metric of resilience during the exposure to perturbations is considered.
 Conclusions. The proposed model architecture and learning algorithm provide absorption of part of the disturbing influence, graceful degradation due to hierarchical classes and adaptive computation, and fast adaptation on a limited amount of labeled data. It is shown that adaptive computation saves up to 40% of resources due to early decision-making in the lower sections of the model, but perturbing influence leads to slowing down, which can be considered as graceful degradation. A multi-section structure trained using knowledge self-distillation principles has been shown to provide more than 5% improvement in the value of the integral mectric of resilience compared to an architecture where the decision is made on the last layer of the model. It is observed that the dimensionality of the feature space noticeably affects the resilience to adversarial attacks and can be chosen as a tradeoff between resilience to perturbations and efficiency without perturbations.