Machine perception is one of the most lucrative domains in the modern landscape and one of the most challenging tasks embodied by this domain is analyzing and interpreting images. Image recognition has seen several advancements over the years such as the introduction of pre-trained models that have expunged the complexity associated with developing high performance deep neural networks. In this paper, we have proposed several eminent pre-trained models that have the ability to categorize dogs as per their breeds. The main objective of this paper is to present a fair comparison of the proposed models and establish the nonpareil model on the basis of comparison metrics such as accuracy, validation accuracy, time requirements, precision and recall scores of the model. It was observed that ResNet 152V2 performed the best with respect to the accuracy, precision and recall scores, Inception-ResNet gave the best validation accuracy and NASNet-Mobile had the highest efficiency albeit with inferior performance in accordance with the other evaluation metrics. 

Analysis of Deep Pre-trained Models for Computer Vision Applications: Dog Breed Classification

Out of all cancers among female patients, breast cancer is on top in terms of prevalence. Many deep learning approaches have been investigated for the classification of histopathological images of breast cancer, however most tend to be either too complex or too large to adopt at a clinical level. In this study we propose a lightweight network, based on knowledge distillation, which performs almost the same as the teacher network but with significantly fewer parameters. Our student network consists of inverted residual blocks of MobileNetV2 and the ghost module. We performed our experiments on BreakHis, BACH and Kaggle breast cancer histopathological imaging datasets. The results show that different versions of our proposed lightweight student architecture perform with similar accuracy levels compared with the teacher network, while using significantly fewer parameters.

Lightweight Deep Network for the Classification of Breast Cancer Histopathological Images

Neural network-based strategies for automatically diagnosing of COVID-19 from X-ray images utilizing different feature extraction algorithms

Cancer is a rampant phenomenon caused by uncontrollable cells that grow and spread throughout the body. Invasive Ductal Carcinoma 1 is the most common type of breast cancer, which can be fatal for females if not detected early. As a result, prompt diagnosis is critical to maximizing surveillance rates and, in the meantime, minimizing long-term mortality rates. Nowadays, modern computer vision and deep learning techniques have transformed the medical image analysis arena. Computer vision application in medical image analysis has provided us with remarkable results, enhanced accuracy, and reduced costs. The main purpose of designing a new algorithm to detect unusual patches of breast images, was to acquire both high accuracy and low computational cost, simultaneously. Therefore, a novel architecture has been designed by utilizing Xception and MobileNetV2.This new algorithm achieves 93.4% balanced accuracy and 94.8% for F1-Score, which outperforms previously published algorithms for identifying IDC histopathology images that use deep learning techniques.

Designing an Improved Deep Learning-Based Classifier for Breast Cancer Identification in Histopathology Images

COVID-19 has adversely affected humans and societies in different aspects. Numerous people have perished due to inaccurate COVID-19 identification and, consequently, a lack of appropriate medical treatment. Numerous solutions based on manual and automatic feature extraction techniques have been investigated to address this issue by researchers worldwide. Typically, automatic feature extraction methods, particularly deep learning models, necessitate a powerful hardware system to perform the necessary computations. Unfortunately, many institutions and societies cannot benefit from these advancements due to the prohibitively high cost of high-quality hardware equipment. As a result, this study focused on two primary goals: first, lowering the computational costs associated with running the proposed model on embedded devices, mobile devices, and conventional computers; and second, improving the model's performance in comparison to previously published methods (at least performs on par with state-of-the-art models) in order to ensure its performance and accuracy for the medical recognition task. This study used two neural networks to improve feature extraction from our dataset: VGG19 and ResNet50V2. Both of these networks are capable of providing semantic features from the nominated dataset. To this end, An alternative network was considered, namely MobileNetV2, which excels at extracting semantic features while requiring minimal computation on mobile and embedded devices. Knowledge distillation (KD) was used to transfer knowledge from the teacher network (concatenated ResNet50V2 and VGG19) to the student network (MobileNetV2) to improve MobileNetV2 performance and to achieve a robust and accurate model for the COVID-19 identification task from chest X-ray images.

Designing an Improved Deep Learning-based Model for COVID-19 Recognition in Chest X-ray Images: A Knowledge Distillation Approach

Medical image analysis techniques have been helpful for rapidly identifying COVID-19. Due to the difficulty of identifying critical visual features for many patients, machine learning techniques are promising for diagnosing infection. Given the variety of approaches to using convolutional-based architectures for medical image analysis tasks, we used the most well-known and powerful deep neural networks to provide a concise and reliable source of information on transfer learning-based methods, paving the way for future research. The article describes an investigation into the detection of COVID-19 in medical chest X-ray images using 27 pre-trained convolutional neural networks. This study utilized a binarized version of a few publicly available datasets from COVID-19 patients and healthy controls. The results of the 5-fold cross-validation demonstrate that architectures such as EfficientNet networks, MobileNet, Inception-ResNetV2, and NasNet-Large perform significantly better on the aforementioned binary pattern recognition task.

An Investigation on Transfer Learning for Classification of COVID-19 Chest X-Ray Images with Pre-trained Convolutional-based Architectures

This paper proposes a novel approach to controller design for an MR-damped vehicle suspension system. This approach is predicated on the premise that the optimal control strategy can be “learned” through real-world or simulated experiments utilizing a reinforcement learning algorithm with continuous states/actions. The sensor data is fed into a Twin Delayed Deep Deterministic Policy Gradient (TD3) algorithm, which generates the actuation voltage required for the MR damper. The resulting suspension working space (displacement), sprung mass acceleration, and dynamic tire load are calculated using a quarter vehicle model incorporating the modified Bouc-Wen MR damper model. Deep RL’s reward function is based on sprung mass acceleration. The proposed approach outperforms traditional suspension control strategies regarding ride comfort and stability, as demonstrated by multiple simulated experiments.

A Deep Reinforcement Learning-Based Controller for Magnetorheological-Damped Vehicle Suspension

In this paper, an adaptive control scheme based on using neural networks is designed to guarantee the desired behavior of a micro-robot which is equipped with vibrating actuators and follows the principle of slip-stick movement. There are two tiny shaking motors which have been utilized to run the micro-class robotic system. Dynamic modeling equations are expressed by considering the spring coefficient of the bases. After that, the effect of the spring on the foundations was investigated. In addition to designing neural-based controller, an AI-based system identifier has been developed to help the controller update its parameters and achieve its desired targets. Using this method, several specific paths for the movement of this micro robot are simulated. Based on the simulation results, the proposed controlling strategy guarantees acceptable performance for tracking different paths due to plotted near-zero errors and handles the nonlinear behavior of the micro-robot system. micro-class its but increasing its accuracy of motion. In the control simulations for several different paths for this micro-robot movement have been performed first using the intelligent PID control method. Following that, the Intelligent-Adaptive PID controller was described, and its application was examined. The simulation results indicate an acceptable performance of the robotic system in tracking tasks.

Amirreza BabaAhmadi

Papers

Designing an Improved Deep Learning-Based Classifier for Breast Cancer Identification in Histopathology Images

Designing an Improved Deep Learning-based Model for COVID-19 Recognition in Chest X-ray Images: A Knowledge Distillation Approach

An Investigation on Transfer Learning for Classification of COVID-19 Chest X-Ray Images with Pre-trained Convolutional-based Architectures

A Deep Reinforcement Learning-Based Controller for Magnetorheological-Damped Vehicle Suspension

Designing an AI-Based Adaptive Controller Augmented with a System Identifier for a Micro-Class Robot Equipped with a Vibrating Actuator