The advent of Tiny Machine Learning (TinyML) has unlocked the potential to deploy machine learning models on resource-constrained edge devices, revolutionizing real-time monitoring in Internet of Medical Things (IoMT) applications. This study introduces a novel approach to real-time electrocardiogram (ECG) anomaly detection by integrating TinyML with edge Artificial Intelligence (AI) on low-power embedded systems. We demonstrate the feasibility and effectiveness of deploying optimized models on edge devices, such as the Raspberry Pi and Arduino, to detect ECG anomalies, including arrhythmias. The proposed workflow encompasses data preprocessing, feature extraction, and model inference, all executed directly on the edge device, eliminating the need for cloud resources. To address the constraints of memory and power consumption in wearable devices, we applied advanced optimization techniques, including model pruning and quantization, achieving an optimal balance between accuracy and resource utilization. The optimized model achieved an accuracy of 92.3% while reducing the power consumption to 0.024 mW, enabling continuous, long-term health monitoring with minimal energy requirements. This work highlights the potential of TinyML to advance edge AI for real-time medical applications.

Improving agricultural productivity is essential due to rapid population growth, making early detection of crop diseases crucial. Although deep learning shows promise in smart agriculture, practical applications for identifying wheat diseases in complex backgrounds are limited. In this paper, we propose CropNet, a hybrid method that utilizes Red, Green, and Blue (RGB) imaging and a transfer learning approach combined with shallow convolutional neural networks (CNN) for further feature refinement. To develop our customized model, we conducted an extensive search for the optimal deep learning architecture. Our approach involves freezing the pre-trained model for feature extraction and adding a custom trainable CNN layer. Unlike traditional transfer learning, which typically uses trainable dense layers, our method integrates a trainable CNN, deepening the architecture. We argue that pre-trained features in transfer learning are better suited for a custom shallow CNN followed by a fully connected layer, rather than being fed directly into fully connected layers. We tested various architectures for pre-trained models including EfficientNetB0 and B2, DenseNet, ResNet50, MobileNetV2, MobileNetV3-Small, and Inceptionv3. Our approach combines the strengths of pre-trained models with the flexibility of custom architecture design, offering efficiency, effective feature extraction, customization options, reduced overfitting, and differential learning rates. It distinguishes itself from classical transfer learning techniques, which typically fine-tune the entire pre-trained network. Our aim is to provide a lightweight model suitable for resource-constrained environments, capable of delivering outstanding results. CropNet achieved 99.80% accuracy in wheat disease detection with reduced training time and computational cost. This efficient performance makes CropNet promising for practical implementation in resource-constrained agricultural settings, benefiting farmers and enhancing production.

The number of connected devices or Internet of Things (IoT) devices has rapidly increased. According to the latest available statistics, in 2023, there were approximately 17.2 billion connected IoT devices; this is expected to reach 25.4 billion IoT devices by 2030 and grow year over year for the foreseeable future. IoT devices share, collect, and exchange data via the internet, wireless networks, or other networks with one another. IoT interconnection technology improves and facilitates people’s lives but, at the same time, poses a real threat to their security. Denial-of-Service (DoS) and Distributed Denial-of-Service (DDoS) attacks are considered the most common and threatening attacks that strike IoT devices’ security. These are considered to be an increasing trend, and it will be a major challenge to reduce risk, especially in the future. In this context, this paper presents an improved framework (SDN-ML-IoT) that works as an Intrusion and Prevention Detection System (IDPS) that could help to detect DDoS attacks with more efficiency and mitigate them in real time. This SDN-ML-IoT uses a Machine Learning (ML) method in a Software-Defined Networking (SDN) environment in order to protect smart home IoT devices from DDoS attacks. We employed an ML method based on Random Forest (RF), Logistic Regression (LR), k-Nearest Neighbors (kNN), and Naive Bayes (NB) with a One-versus-Rest (OvR) strategy and then compared our work to other related works. Based on the performance metrics, such as confusion matrix, training time, prediction time, accuracy, and Area Under the Receiver Operating Characteristic curve (AUC-ROC), it was established that SDN-ML-IoT, when applied to RF, outperforms other ML algorithms, as well as similar approaches related to our work. It had an impressive accuracy of 99.99%, and it could mitigate DDoS attacks in less than 3 s. We conducted a comparative analysis of various models and algorithms used in the related works. The results indicated that our proposed approach outperforms others, showcasing its effectiveness in both detecting and mitigating DDoS attacks within SDNs. Based on these promising results, we have opted to deploy SDN-ML-IoT within the SDN. This implementation ensures the safeguarding of IoT devices in smart homes against DDoS attacks within the network traffic.

The ability to differentiate between various fish species plays an essential role in aquaculture. It helps to protect their populations and monitor their health situations and their nutrient systems. However, old machine learning methods are unable to detect objects in images with complex backgrounds and especially in low-light conditions. This paper aims to improve the performance of a YOLO v5 model for fish recognition and classification. In the context of transfer learning, our improved model FishDETECT uses the pre-trained FishMask model. Then it is tested in various complex scenes. The experimental results show that FishDETECT is more effective than a simple YOLO v5 model. Using the evaluation metrics Precision, Recall, and mAP50, our new model achieved accuracy rates of 0.962, 0.978, and 0.995, respectively.

Fish play a major role in the human nutritional system, and farmers need to know the accurate prediction of fish weight in order to optimize the production process and reduce costs. However, existing prediction methods are not efficient. The formulas for calculating fish weight are generally designed for a single species of fish or for species of a similar shape. In this paper, a new hybrid method called SFI-XGBoost is proposed. It combines the VIF (variance inflation factor), PCC (Pearson’s correlation coefficient), and XGBoost methods, and it covers different fish species. By applying GridSearchCV validation, normalization, augmentation, and encoding techniques, the obtained results show that SFI-XGBoost is more efficient than simple XGBoost. The model generated by our approach is more generalized, achieving accurate results with a wide variety of species. Using the r2_score evaluation metric, SFI-XGBoost achieves an accuracy rate of 99.94%.

This paper presents a proposed artificial intelligence (AI) framework used to detect attacks in an IoT ecosystem. The proposed framework is an intrusion detection system (IDS) using a machine learning strategy to monitor the IoT system and detect malicious and suspicious activity. In this work, we used supervised machine learning (ML) method to increase detection accuracy and minimize data processing time. Four classification algorithms are used to evaluate the proposed model, namely Random Forest (RF), Support Vector Machines (SVM), k Nearest Neighbors (kNN) and Gaussian Naïve Bayes (GNB) algorithm. Experimental results showed that KNN performs best with 97.5% accuracy and fastest training times with 0.03 seconds (i.e. the training times are the CPU time to build the model). At the end, we proposed an implementation of the proposed IDS in a real IoT environment.