Failure due to cracks is a major structural safety issue for engineering constructions. Human examination is the most common method for detecting crack failure, although it is subjective and time-consuming. Inspection of civil engineering structures must include crack detection and categorization as a key component of the process. Images can automatically be classified using convolutional neural networks (CNNs), a subtype of deep learning (DL). For image categorization, a variety of pre-trained CNN architectures are available. This study assesses seven pre-trained neural networks, including GoogLeNet, MobileNet-V2, Inception-V3, ResNet18, ResNet50, ResNet101, and ShuffleNet, for crack detection and categorization. Images are classified as diagonal crack (DC), horizontal crack (HC), uncracked (UC), and vertical crack (VC). Each architecture is trained with 32,000 images equally divided among each class. A total of 100 images from each category are used to test the trained models, and the results are compared. Inception-V3 outperforms all the other models with accuracies of 96%, 94%, 92%, and 96% for DC, HC, UC, and VC classifications, respectively. ResNet101 has the longest training time at 171 min, while ResNet18 has the lowest at 32 min. This research allows the best CNN architecture for automatic detection and orientation of cracks to be selected, based on the accuracy and time taken for the training of the model.

In recent years, incidents involving unmanned aerial vehicles (UAVs) have increased significantly, raising concerns over security and privacy, especially concerning civilian and military facilities. Vision-based approaches, especially those employing deep convolutional neural networks (DCNNs), show great promise in addressing the need for an accurate and cost-effective UAV detection system. However, DCNNs rely heavily on extensive and well-labeled datasets, which are essential for achieving high accuracy and effectiveness. For drone detection tasks, a dataset of high-resolution images is especially valuable, as it provides more contextual information for DCNNs, enabling more accurate drone detection. This work presents a new drone detection dataset composed of 4K resolution images named University of Engineering and Technology Taxila 4K Anti-UAV (UETT4K Anti-UAV). The dataset is created by obtaining real-world videos of different types of drones in diverse environmental and challenging conditions. A custom dataset of 33601 images, manually annotated with hand-labeled bounding boxes, is created from these videos.We used our dataset to train, validate, and test eight state-of-the-art YOLOv6v3 algorithm models. The test results and qualitative analysis guide selecting the most suitable model for drone detection applications. The proposed extensive 4K dataset offers a valuable resource for effectively training deep learning models to detect drones across diverse conditions accurately.

Rate-Distortion optimization is updated to improve HEVC coding performance for HDR content. The proposed method employs Perceptually Uniform encoding as a distortion measure and shows on average performance improvement of 1.06% for HDR-VDP-2.2 Q factor and 4.02% in HDR-VQM with respect to the reference encoder. Further, with proposed change, better bit budget estimation is achieved showing performance improvement in Rate-Control. In comparison to NRMSE values from 0.5991 to 3.7670 achieved by the reference encoder, the proposed method is able to achieve improved NRMSE values from 0.2896 to 2.7754.