Recent research on single-image super-resolution (SISR) using deep convolutional neural networks has made a breakthrough and achieved tremendous performance. Despite their significant progress, numerous convolutional neural networks (CNN) are limited in practical applications, owing to the requirement of the heavy computational cost of the model. This paper proposes a multi-path network for SISR, known as multi-path deep CNN with residual inception network for single image super-resolution. In detail, a residual/ResNet block with an Inception block supports the main framework of the entire network architecture. In addition, remove the batch normalization layer from the residual network (ResNet) block and max-pooling layer from the Inception block to further reduce the number of parameters to preventing the over-fitting problem during the training. Moreover, a conventional rectified linear unit (ReLU) is replaced with Leaky ReLU activation function to speed up the training process. Specifically, we propose a novel two upscale module, which adopts three paths to upscale the features by jointly using deconvolution and upsampling layers, instead of using single deconvolution layer or upsampling layer alone. The extensive experimental results on image super-resolution (SR) using five publicly available test datasets, which show that the proposed model not only attains the higher score of peak signal-to-noise ratio/structural similarity index matrix (PSNR/SSIM) but also enables faster and more efficient calculations against the existing image SR methods. For instance, we improved our method in terms of overall PSNR on the SET5 dataset with challenging upscale factor 8× as 1.88 dB over the baseline bicubic method and reduced computational cost in terms of number of parameters 62% by deeply-recursive convolutional neural network (DRCN) method.

Single image super-resolution (SISR) aims to reconstruct a high-resolution (HR) image from a low-resolution (LR) image. In order to address the SISR problem, recently, deep convolutional neural networks (CNNs) have achieved remarkable progress in terms of accuracy and efficiency. In this paper, an innovative technique, namely a multi-scale inception-based super-resolution (SR) using deep learning approach, or MSISRD, was proposed for fast and accurate reconstruction of SISR. The proposed network employs the deconvolution layer to upsample the LR image to the desired HR image. The proposed method is in contrast to existing approaches that use the interpolation techniques to upscale the LR image. Primarily, interpolation techniques are not designed for this purpose, which results in the creation of undesired noise in the model. Moreover, the existing methods mainly focus on the shallow network or stacking multiple layers in the model with the aim of creating a deeper network architecture. The technique based on the aforementioned design creates the vanishing gradients problem during the training and increases the computational cost of the model. Our proposed method does not use any hand-designed pre-processing steps, such as the bicubic interpolation technique. Furthermore, an asymmetric convolution block is employed to reduce the number of parameters, in addition to the inception block adopted from GoogLeNet, to reconstruct the multiscale information. Experimental results demonstrate that the proposed model exhibits an enhanced performance compared to twelve state-of-the-art methods in terms of the average peak signal-to-noise ratio (PSNR), structural similarity index (SSIM) with a reduced number of parameters for the scale factor of $2\times$ , $4\times$ , and $8\times$ .

Deep learning (DL) architectures are becoming increasingly popular in modern traffic systems and selfdriven vehicles owing to their high efficiency and accuracy. Emerging technological advancements and the availability of large databases have made a favorable impact on such improvements. In this study, we present a traffic sign recognition system based on novel DL architectures, trained and tested on a locally collected traffic sign database. Our approach includes two stages; traffic sign identification from live video feed, and classification of each sign. The sign identification model was implemented with YOLO architecture and the classification model was implemented with Xception architecture. The input video feed for these models were collected using dashboard camera recordings. The classification model has been trained with the German Traffic Sign Recognition Benchmark dataset as well for comparison. Final accuracy of classification for the local dataset was 96.05% while the standard dataset has given an accuracy of 92.11%. The final model is a combination of the detection and classification algorithms and it is able to successfully detect and classify traffic signs from an input video feed within an average detection time of 4.5fps

Malaria remains a critical global health challenge for doctors and healthcare practitioners, particularly clinicians involved in initial treatment. Inaccurate diagnosis of malaria-infected cells often leads to delayed or inappropriate treatment, increasing the risk of severe complications or death. Traditional microscopic diagnosis is time-consuming and requires expert skills, resulting in variability and inconsistency in results. These challenges are further complicated by the complexity of malaria symptoms, which overlap with other febrile illnesses, making clinical diagnosis unreliable without laboratory confirmation. To address these challenges, this study explores deep-learning-based approaches, particularly leveraging state-of-the-art pretrained convolutional neural network (CNN) models, for automated malaria parasite detection and classification from microscopic blood smear images. Transfer learning is an effective approach to handling issues such as limited labeled data, time-consuming training, and domain-specific variations in medical image classification. By leveraging pretrained models trained on large-scale datasets like ImageNet, transfer learning enables the reuse of learned features, significantly accelerating the adaptation process for malaria detection and other medical imaging tasks. We used eight pretrained models for malaria parasite classification such as VGG16, VGG19, Inception-v3, ResNet-18, ResNet-34, ResNet-50, ResNet-101, and Xception. In particular, ResNet-50 and ResNet-101 achieved accuracies of approximately 89%, respectively, while Xception reached around 88% accuracy. In comparison, VGG-16 achieved a lower overall accuracy of about 80% due to a recall trade-off despite high precision. These metrics highlight meaningful improvements over simpler architectures and validate the efficacy of our transfer learning approach for automated malaria detection. The proposed models were fine-tuned on extensive labeled datasets comprising parasitized and uninfected cells. Quantitative and qualitative evaluations were conducted using metrics such as precision, recall, F1-score, and support. Our experimental results demonstrate that ResNet-50, ResNet-101, and Xception exhibit strong balanced performance with higher accuracy, while VGG-16 shows a trade-off of high precision but lower recall for parasitized cells.