Based on the background of rapid transmission of novel coronavirus and various pneumonia, wearing masks becomes the best solution to effectively reduce the probability of transmission. For a series of problems arising from crowded public places and collective units, where face recognition is difficult to increase target density, a deep convolutional neural network is used for real-time mask detection and recognition.This paper presents the method based on YOLOv5 model for deep learning and mask detection in image recognition as well as a live camera to label the pedestrians without masks in time. This experiment will use LabelImg software to preprocess 5003 images and make lightweight improvements based on the original YOLOv5 model to generate the final face mask recognition model. The Mosaic method is added to merge the images effectively and process the images in batch, and secondly, the GIoU loss function is selected to calculate the bounding box regression loss by comparison, which improves the localization accuracy even more. According to the experimental detection results, analogized with the original model YOLOv5, the recall and accuracy are effectively improved. In this paper, YOLOv3, SSD, Fast-R-CNN detection algorithms are used for comparison, the detection results of this model have a high mAP value which is equal to 92.9, which are higher than the detection results of other models. Real-time target recognition based on this model combined with practical applications can be applied in hospitals and crowd-gathering places to achieve effective reduction of epidemic transmission probability in a short period of time.

Nowadays, chatbots are used only on a few websites. People can’t always adapt to the design of every website, so they need an assistant for that. On every website, a chatbot can find its calling, and it can be used for different needs, and for solving different problems. One of the main problems on the Internet is copyright infringement, copying other people’s sites. Various fake websites copy official sources, and these types of fraud disguise all sorts of links as normal official links, and people often fall for this deception. Such websites may also use chatbots for their evil designs. By spreading scam links using chatbots, scammers can gain access to users’ data, social media accounts, bank card details, and other important data. Therefore, there is a need for a thorough study of chatbots, as well as their correct and effective use. A chatbot is a multifaceted product that can be used in any area of life, from simple information searches to education and medicine. This article will present some types of chatbots, as well as information on where they can be used and for what needs. The purpose of the article is to analyze the effectiveness of chatbots and also offer some methods for determining the effectiveness of using a chatbot. There will be information about the first chatbots, about the trends of our days. Also in this article, a functioning chatbot model will be presented, which will be used in one of the most popular communication platforms, Discord.

Research and analysis of demographic processes play an important role in many areas. For this, the population size and key factors from 1994 to 2019 were selected on the statistical website of the Republic of Kazakhstan. Demographics were population size, fertility, mortality, divorce, and migration. The factors of the standard of living were the number of unemployed and the average monthly salary, while the medical factors were the hospital organizations, the number of hospital beds and the number of doctors of all specialties. In the course of regression analysis, a correlation was obtained and multicollinear factors were identified. We used four different machine learning models from the Scikit-Learn library to generate population estimates. Regression models were evaluated using the quality score. As a result, linear regression and random forest models performed well.

The process of sublimation drying of berries provided, this is the drying method along the berry the humidity is reduced to 2-5%. The dryer is conditional consists of two blocks - refrigeration and sublimation unit. At room temperature the product is placed in the drying chamber. Equipment after connection, vacuum pump in the chamber reduces the pressure to 10-30 Pa. When a vacuum is created and free product due to evaporation of moisture leads to freezing. Most of the moisture is converted into ice crystals, and then the sublimation process begins. This equipment uses 404 a freon. Temperature to register 8 channel OVEN TRM138 RS-485 temperature meter was used. Of the product to measure weight every 30 minutes removed from the chamber Vibra HT 224RCE was weighed on an analytical balance. The experiment was performed with three types of berries: strawberries; currants and raspberries. Experience during the temperatures on the surface of the berries, inside the chamber and the mass of the product were measured. Duration of drying as a criterion of effective mode, energy expenses and quality indicators of dried berries were determined. The effective sublimation time was 4.5 hours. Inside the chamber at a pressure of (20 ± 5) Pa, at a temperature of (-42 ± 2) 0С, general freezing and sublimation time not more than 5 hours, the total drying time is 8 hours. The initial weight of strawberries decreased by 36%, currants by 28% and raspberries by 27.4%. Drying should be stopped when the chamber temperature reaches 40 ° C.

The object of the study is the accuracy of diabetic retinopathy diagnosis based on retinal images. This study investigates convolutional neural network (CNN) models for the automatic detection of diabetic retinopathy (DR) from retinal images. The main problem lies in the insufficient effectiveness of basic CNN models in recognizing DR stages on fundus images. The core problem addressed is the suboptimal performance of baseline CNNs in identifying DR stages from medical imagery. To solve this, two CNN architectures were thoroughly evaluated: a baseline model and an enhanced model integrating advanced preprocessing techniques such as image resizing (256 × 256 and 512 × 512), the image normalization, and data augmentation methods. The enhanced model outperformed the original, achieving a validation accuracy of 91% compared to 88% for the baseline, and demonstrating reduced loss during both training and validation. This improvement is attributed to the optimized input image quality and increased variability in the training set, which enhanced the model’s ability to generalize and avoid overfitting. A distinctive feature of the results lies in the synergy between preprocessing and CNN architecture, which enabled significantly improved classification performance even under hardware constraints. These limitations suggest that further gains are possible with extended computational resources and access to larger datasets. The practical applicability of the findings is evident in the potential deployment of such models in clinical screening systems to support early and accurate DR diagnosis. The models were trained on a proprietary dataset of expert-labeled, high-resolution retinal images, similar in format to EyePACS and APTOS, though not publicly available due to ethical considerations