Video is considered as one of the most useful and important forms of multimedia data, that is usually used in several applications. Despite its importance, video indexing and retrieval becomes a challenging task. In order to reduce the amount of data and keep only relevant frames, keyframe extraction becomes necessary in a content‐based video retrieval (CBVR) system. In this paper, a keyframe extraction method is proposed based on the face image quality for video surveillance systems. Data is reduced by rejecting frames without faces. Then, face images are clustered by identity. After that, a set of candidate frames is selected to be proceeded. The face quality assessment is based on four metrics including pose estimation, sharpness, brightness and resolution, and the frame with the best face quality is considered as a keyframe. Experimental tests were carried on several datasets in order to prove the efficiency of authors' method compared with state‐of‐the‐art approaches.

This article introduces a comprehensive dataset designed to facilitate smart harvesting applications for Deglet Nour date palms, focusing on two primary computer vision tasks: branch detection and optimal cutting point localization. Collected from various oases within the Kebili governorate of southern Tunisia during the peak harvesting season of 2023, the dataset consists of 5530 images annotated for object detection and 387 images precisely annotated with keypoints. Captured using a combination of DSLR and smartphone cameras, the dataset captures real- world agricultural complexities, including varied lighting conditions and occlusions, which were covered and uncorvered date bunches. Annotations, performed manually using Roboflow, are provided in widely used YOLO and COCO formats to ensure compatibility and facilitate widespread adoption. By offering structured, high quality annotated images, this dataset supports the robust development, training, and evaluation of advanced computer vision and robotic harvesting systems for precision agriculture.

Face recognition (FR) in unconstrained conditions remains an open research topic and an ongoing challenge. The facial images exhibit diverse expressions, occlusions, variations in illumination, and heterogeneous backgrounds. This work aims to produce an accurate and robust system for enhanced Security and Surveillance. A parallel hybrid deep learning model for feature extraction and classification is proposed. An ensemble of three parallel extraction layer models learns the best representative features using CNN and RNN. 2D LBP and 3D Mesh LBP are computed on face images to extract image features as input to two RNNs. A stacked autoencoder (SAE) merged the feature vectors extracted from the three CNN-RNN parallel layers. We tested the designed 2D/3D CNN-RNN framework on four standard datasets. We achieved an accuracy of $98.9\%$. The hybrid deep learning model significantly improves FR against similar state-of-the-art methods. The proposed model was also tested on an unconstrained conditions human crowd dataset, and the results were very promising with an accuracy of $95\%$. Furthermore, our model shows an 11.5% improvement over similar hybrid CNN-RNN architectures, proving its robustness in complex environments where the face can undergo different transformations.

COVID-19 is a disease that infects people and quickly isolates the entire world. The new variants of COVID-19 continue to cause high mortality rates. Therefore, many scientists worldwide still are looking for a solution to quickly and accurately detect COVID-19. This paper aims to detect COVID-19 using chest CT-Scan and Chest X-ray images. In this work, we design a new bimodal convolutional neural network (CNN) that requires two inputs. The first modality is a chest CT-Scan image segmented by a U-Net deep learning technique to detect infected areas in the lung. The second input is a Chest X-ray image. The proposed CNN combines the features extracted from these two images. Feature extraction is performed on these two input images in two parallel feature extraction layers. The extracted feature vectors will be combined by a perceptron attention mechanism and taken as input to fully connected layers to classify the patient as COVID-19, non-COVID, and pneumonia. The results have shown that the newly designed CNN outperforms other similar state-of-the-art methods especially in distinguishing between pneumonia and COVID-19 cases. The proposed CNN has achieved 98.79% classification accuracy and 43.20% loss. The proposed framework could be particularly beneficial in telemedicine, enabling remote diagnosis in areas with limited access to medical specialists.