Human facial emotion recognition (FER) has attracted the attention of the research community for its promising applications. Mapping different facial expressions to the respective emotional states are the main task in FER. The classical FER consists of two major steps: feature extraction and emotion recognition. Currently, the Deep Neural Networks, especially the Convolutional Neural Network (CNN), is widely used in FER by virtue of its inherent feature extraction mechanism from images. Several works have been reported on CNN with only a few layers to resolve FER problems. However, standard shallow CNNs with straightforward learning schemes have limited feature extraction capability to capture emotion information from high-resolution images. A notable drawback of the most existing methods is that they consider only the frontal images (i.e., ignore profile views for convenience), although the profile views taken from different angles are important for a practical FER system. For developing a highly accurate FER system, this study proposes a very Deep CNN (DCNN) modeling through Transfer Learning (TL) technique where a pre-trained DCNN model is adopted by replacing its dense upper layer(s) compatible with FER, and the model is fine-tuned with facial emotion data. A novel pipeline strategy is introduced, where the training of the dense layer(s) is followed by tuning each of the pre-trained DCNN blocks successively that has led to gradual improvement of the accuracy of FER to a higher level. The proposed FER system is verified on eight different pre-trained DCNN models (VGG-16, VGG-19, ResNet-18, ResNet-34, ResNet-50, ResNet-152, Inception-v3 and DenseNet-161) and well-known KDEF and JAFFE facial image datasets. FER is very challenging even for frontal views alone. FER on the KDEF dataset poses further challenges due to the diversity of images with different profile views together with frontal views. The proposed method achieved remarkable accuracy on both datasets with pre-trained models. On a 10-fold cross-validation way, the best achieved FER accuracies with DenseNet-161 on test sets of KDEF and JAFFE are 96.51% and 99.52%, respectively. The evaluation results reveal the superiority of the proposed FER system over the existing ones regarding emotion detection accuracy. Moreover, the achieved performance on the KDEF dataset with profile views is promising as it clearly demonstrates the required proficiency for real-life applications.

The malignant neoplastic malady known as cancer appears to exhibit a significantly elevated rate of mortality owing to its virulence and pronounced propensity for metastasis. To augment the diagnostic efficacy, research endeavors have been undertaken utilizing complex deep learning architectures. However, the performance of these efforts remains circumscribed by smaller dataset size, quality of the data, the interclass variations present between lung adenocarcinoma and lung squamous cell carcinoma, and the complexity of deploying to mobile devices and failure to address both image and patient level accuracy measurements. To surmount these obstacles, the present study proposes a stage-based method for enhancing the images, in conjunction with utilizing a global context attention-guided convolutional neural network that effectively captures both channel and spatial information and semantic information extracted from the input image. Implementing the proposed methodology increased total image level accuracy to 99.76% and a patient level accuracy of 96.5%, a metric that has yet to be previously quantified. The addition of the global context attention module decreases the model’s parameter count by 0.47 million, reduces the computational costs by saving 10.54 million floating point operations per second (FLOPs) and 10.72 million multiply-accumulate operations (MACs), and results in a 0.03s improvement in inference time. Furthermore, this module enhances both image level and patient level accuracy, boosting them by 2.84% and 3.17%, respectively, compared to using only the convolutional block attention module in the baseline convolutional neural network. Consequently, this modification renders the model highly suitable for deployment on mobile devices due to its adaptability. Our findings provide supporting evidence for the potential of this method to serve as a noninvasive screening tool capable of reliably classifying lung and colon cancer subtypes.

Breast cancer is the leading cause of death among women. Timely diagnosis improves patient survival rates. However, classifying breast cancer histopathological slides using deep learning faces challenges. These challenges include limited datasets, multiple magnification factors, interference from irrelevant features, complex variations between classes, and issues with model interpretability. To address these challenges, we introduce a new attention-guided deep broad convolutional neural network (ADBNet). The ADBNet has a modified convolutional block attention module that focuses on selective features while suppressing irrelevant ones. It also has a deep broad block that enhances the network’s resilience to various magnification factors. Additionally, we employ a generative adversarial network combined with diffusion to expand the dataset with expertly validated images. This enriches the dataset for classifying similar cancer subtypes. The ADBNet achieves remarkable image level recognition accuracy: 99.33% at 40x magnification, 99.52% at 100x, 99.13% at 200x, and 99.06% at 400x magnification. It also attains high patient-level recognition accuracy: 99.05% at 40x, 99.15% at 100x, 99.03% at 200x, and 98.60% at 400x magnification. Impressively, for magnification-independent classification, our approach achieves 99.36% image level and 99.28% patient-level recognition accuracy. We evaluate the proposed model on publicly available datasets, including breast histopathology images, LC25000, and Cifar-10. The results surpass the performance of existing methods. This reinforces the efficacy and potential of the ADBNet for classifying breast cancer histopathological images.

Trash production and disposal have emerged as serious issues for underdeveloped nations as their populations have swelled. As manual classification can be both time-consuming and potentially dangerous, therefore, nowadays, it is increasingly being replaced by automated methods. Recent advances in AI and deep learning have allowed for significant advancements in trash detection and classification systems. Due to the lack of a comprehensive trash detection dataset tailored to Bangladesh, we set out to collect data that would accurately portray the complexity of Bangladesh’s scenario while also incorporating openlittermap. In this study, we employ a deep learning model known as YOLOv5. Several variants of the YOLOv5 model are used and assessed with both the freshly minted dataset and the already existing benchmark datasets. Simulation results indicate that the finetuned YOLOv5 model outperforms existing models in terms of mean average precision (mAP) and F1-score. On the Bangladeshi dataset, the model shows an mAP of 34.3% and an F1-score of 43.7%. The mAP and F1-score provide a holistic evaluation of YOLOv5’s object recognition accuracy, localization, and precision-recall balance. By incorporating the additional data from openlittermap into the new dataset, the mAP is increased to 45.4%. In addition, for some variants of YOLOv5, the suggested model produces greater mAP than the current literature on both the TACO and PlastOpol datasets. The model also achieves an mAP of 84.4% and an F1-score of 78.2% in single-class detection experiments with the newly created dataset. This is because concentrating on just one class helps eliminate class ambiguity, improves localization accuracy, and mitigates class imbalance.

In this paper, we design and implement a multiantenna configured secure millimeter-wave (mmWave) CP-less multi-user orthogonal chirp division multiplexing (OCDM) transceiver. Our proposed simulated system emphasizes more applicable performance matrices for a typically assumed case of four users and a passive eavesdropper for audio data transmission. It introduces a four-dimensional hyperchaotic system-based encryption algorithm to enhance physical layer security (PLS). In addition, low density parity check (LDPC), TURBO, (3, 2) single parity check (SPC), and repeat and accumulate (RA) channel coding with Cholesky decomposition based zero-forcing (CD-ZF) and minimum mean square error (MMSE) signal detection techniques for a better bit error rate (BER) are also implemented. The simulation results signify the effectiveness of the proposed system in terms of PLS enhancement with low correlation coefficients (14.62%, 7.61%, 13.61%, and 15.39% for users 1, 2, 3, and 4, respectively), an achievable secrecy rate with a low signal-to-interference and noise ratio (SINR) of the passive eavesdropper, an achievable out-of-band (OOB) power emission of 341 dB, an estimated average short-time Fourier transform (STFT) spectral power difference of 7.68 dB and estimated peak-to-average power ratios (PAPRs) ranging from 7 to 7.5 dB at a CCDF of 1×10 ^-3 for different ground transmitting channels. At an identical signal-to-noise ratio of 17 dB, all four users achieve a BER of 1×10 ^-4 under RA channel coding, CD-ZF, and 16-QAM digital modulation.