The VGG-BiLSTM-CTC model has succeeded in many languages, yet Russian handwriting recognition is understudied despite its demand. This research adapts the model to Russian through transfer learning to address this gap.

This paper explores the way AI technology systematically promotes the digital transformation of supply chains in Chinese industrial enterprises. It constructs a three-tier organizational economic mechanism model comprising "driving-coresupporting" layers. In this study, the mechanism was confirmed to be effective and significantly improve decision-making optimization, demand ...

This paper compares YOLOv9–YOLOv12 object detection models on a custom non-open dataset (6,500 images, 11 classes), all trained for 300 epochs under identical settings. Results show YOLOv12 achieves the highest mAP@0.5 (0.678), outperforming YOLOv9 (0.648), YOLOv10 (0.603), and YOLOv11 (0.649). Architectural optimizations and training stability drive performance disparities, with ...

This paper presents a framework that integrates Goal-Oriented Bodily Kinesthetic Intelligence (BKI) with disease-oriented walking feature mining using plantar insole sensor signals. By combining BKI’s adaptive motor-control mechanisms with a Neurological Disease Detection Feature Mining Method (NDFMM), the system extracts interpretable gait features and links them to decision ...

This paper presents an intelligent trading framework based on large language models. This system integrates the semantic reasoning of LLMs with classic statistical models, using LLMs to analyze news, sentiment, and macro reports to generate signals. These signals are combined with historical data and then drive the risk control module. The trading system replaces human analysts, ...

This survey reviews classical image compression methods for CPU environments, excluding deep learning techniques. It covers pixel-domain predictive coding (DPCM, JPEG-LS), frequency-domain transform coding (JPEG with DCT, JPEG2000 with wavelets), and modern image formats (WebP, HEIF/AVIF, JPEG XL). The focus is on their operating principles, computational complexity, artifacts, ...

This work contrasts traditional parametric walking models, which handle simple gait well but fail on nonlinear or disease-affected patterns, with a Non-Parametric Walking Model (NPWM) built from plantar insole time–spatial features. The Neurological Disease Detection Feature Mining Method (NDFMM) identifies goal-oriented walking patterns through sorted-metrics–based selection. ...

The article presents the results of the development of the Confinity digital platform for organizing and supporting scientific events. The developed solution is designed to increase the efficiency of scientific communications and reduce the administrative burden on the organizers. The created payment form allows you to automate the processes of submitting applications, content ...

In visual navigation and positioning tasks, agents often fail to simultaneously observe two distant feature points due to the limited field of view (𝐹𝑂𝑉) of the camera, leading to angle measurement failure. This paper proposes a chain tracking measurement method, this method does not rely on additional hardware, utilizing only natural feature points in the image sequence for ...

This paper considers a structure for onboard communication equipment. Its purpose is to collect and process information about the road situation and transmit the data for further processing by artificial intelligence (AI) systems, enabling advanced driver-assistance and traffic optimization solutions.