The attendance record monitors the student's participation in university academic activities, reflecting the commitment to their professional training. However, traditional systems require moderate time to perform this activity and can be susceptible to fraud and errors. In today's technological landscape, facial recognition has become an effective solution to problems in various fields. Currently, all university professors own smartphones. Considering this advantage, this article proposes to develop a mobile application for the registration of academic attendance using advanced artificial intelligence technologies such as Multitasking Cascade Convolutional Networks (MTCNN) in facial detection, MobileFaceNet in facial feature extraction (facial vector) and the Euclidean distance function in the calculation of similarity between obtained vectors. MobileFaceNet was evaluated in Python, using a personalized dataset of top-level students of the Software career of the Universidad Técnica del Norte, achieving an accuracy of 98.9% and 99.4% in LWF. The models were then integrated into a mobile app developed with Android Studio. Finally, the time required to register attendance was compared using the university academic platform (SIIU) and the facial recognition mobile application. The benchmarking showed a 24-second reduction of 33% in attendance registration time.

Weed identification and quantification are processes that are usually manual, subjective, and error-prone. Weeds compete with crops for nutrients, minerals, physical space, sunlight, and water. Thus, weed identification is a crucial component of precision agriculture for autonomous removal and site-specific treatments, efficient weed control, and sustainability. Convolutional Neural Networks (CNNs) are very common in weed identification. This work implemented CNN models for semantic segmentation based on the U-Net architecture for automatically segmenting and quantifying weeds in potato crops using RGB images acquired by a drone at 9–10 m height, flying at 1 m/s. Remote sensing images are affected by factors that degrade image quality and the model’s accuracy. Five U-Net variants were evaluated: the original U-Net, Residual U-Net, Double U-Net, Modified U-Net, and AU-Net. The models were trained using the TensorFlow/Keras frameworks on Google Colab Pro+, following the Knowledge Discovery in Databases (KDD) methodology for image analysis. Each model was trained using a diverse custom dataset in uncontrolled environments, considering six classes: background, Broadleaf dock (Rumex obtusifolius), Dandelion (Taraxacum officinale), Kikuyu grass (Cenchrus clandestinum), other weed species, and the crop potato (Solanum tuberosum L.). The models’ segmentation was widely assessed using Mean Dice Coefficient, Mean IoU, and Dice Loss metrics. The results showed that the Residual U-Net model performed the best in multi-class segmentation, achieving a Mean IoU of 0.8021, a performance comparable to or superior to that reported by other authors. Additionally, a Student’s t-test was applied to complement the data analysis, suggesting that the model is reliable for weed quantification.

The All-Pairs Shortest Paths (APSP) algorithm is crucial for a wide range of computing applications that necessitate the efficient calculation of minimum distances between all pairs of nodes in a graph. High-performance computing (HPC) has made significant progress in response to the demand to solve complex computing problems that process large amounts of data. The cubic complexity of the Floyd-Warshall algorithm makes its execution a computational challenge with large graphs, limiting its applicability on conventional platforms. This research addresses this problem by evaluating optimized and parallelized implementations in HPC architectures. The study aims to evaluate the performance of the APSP algorithm by parallel implementation using Intel DPC++ on heterogeneous high-performance architectures, including CPU, GPU, and FPGA. The Intel OneAPI platform was used for the implementation and execution of the algorithm on Intel Xeon Gold 6128 CPU-3.40 GHz processors, Intel Data Center GPU Max 1100, and FPGA Emulation Device processors. The solution was evaluated considering quantitative and statistical metrics, obtaining a significant improvement in the performance of the algorithm for large graphs, with 700x speedup and 70% efficiency in GPU architectures with 65536 nodes and 1024 parallelism subblocks. This study confirms the feasibility of implementing efficient and portable HPC solutions using DPC++ for massive data processing in heterogeneous architectures, offering a scalable and effective solution for modern computational challenges.