This paper presents a comparative study of key performance metrics for OCR engines in Bangla Language Processing, focusing on PyTesseract and EasyOCR. The models were benchmarked on the “Bangla-CrossHair” dataset, which includes a mix of Bangla text, handwritten characters, and words, tested across blurred, clear, torn, and tilted images. EasyOCR outperformed PyTesseract in several scenarios, though PyTesseract was faster overall. Since OCR engines are generally pre-trained, metrics like training and validation accuracy were not applicable to PyTesseract. The comparison metrics included Character and Word Level Accuracy, Levenshtein Distance, Character and Word Error Rate, Precision, Recall, and F1-Score.

This paper presents a novel real-time GPS (Global Positioning System) and PTS (Passenger Tracking System) architecture by utilizing generative AI. It covers simulation, hardware setup, and application development. A Node MCU Development Board has been utilized to establish a server paired with the GPS and GSM/GPRS modules with a GPS antenna. A sensor hub with PIR, IR, RTS and ultrasonic sensors is built, powered by the ATMega328P. The simulations were conducted in Proteus. A thorough study on the data transmission delay for this system has been added, providing low-level insights into the processing time, transmission time, server response time, network latency, and specifications for the power supply unit. For the application, the frontend harnessed Flutter/DART, Google Maps, and the OpenAI API, while the backend utilized Firebase for making API calls between the ESP8266 web server and the application including user authentication, security and database management on the backend.

This paper describes the development and assessment of a voice-activated assistant system that is intended to address both general knowledge queries and departmental information retrieval. The system integrates Python-based libraries for speech recognition, natural language processing, and text-to-speech conversion, utilizing the Raspberry Pi 4 as the primary hardware. The system utilizes a hybrid mechanism that integrates OpenAI GPT-3.5-Turbo API with a custom departmental database to process voice commands and provide precise responses. YouTube search and playback, Wikipedia integration, voice-activated web browsing, and current affairs updates via the News API are among the additional features. It is illustrated that the system is effective in a variety of acoustic environments, which underscores its potential to improve user interaction in institutional and academic environment.

This paper utilizes MATLAB's Image Processing Toolbox for detecting the license plate region and extracting characters, which are classified using a custom-trained ResNet50 on a dataset of Bangla and English characters, numbers, and noise categories. The model processes vehicle images, identifies license plates, and converts outputs to strings saved in an Excel sheet. The aim is to assess the performance of the Linear SVM classifier with ResNet50 for Bangla OCR. The model achieved 97.57% training accuracy and 99.2% testing accuracy, with an error rate and false positive rate below 0.02%, and 100% F1 scores and Matthews Correlation Coefficient across all categories.

The research paper "Development of Automated Protection and Monitoring System for Poor Railway Infrastructure" was accepted and presented at "PPG Institute of Technology, Coimbatore, India's" "3rd International Conference on Communication, Computing and Electronics Systems (ICCCES 2021)."Scopus Indexed Series (Springer) "Lecture Notes in Electrical Engineering" will publish the study article. This research is based on developing different sub-models for maximizing railway protection. The main motive of developing the railway protection and monitoring system is to help the railway coordinators and locomotive drivers to ensure proper safety to the railway passengers. Three individual sub-models are accumulated to develop the entire model. Firstly, in order to minimize level crossing accidents, the gateman will be notified about the exact arrival time of the train. In addition, by using sensors, the crossing gate will be automatically closed if any train crosses the level crossing. Moreover, the proposed model can scan the train routes in order to avoid collision accidents with other trains or any obstacles across the route. This entire model used many modern technologies, and it is almost errorless. The countries which possess poor rail infrastructure can implement this cheap and reliable protection and monitoring system in order to avoid major rail accidents. 

The research paper "Development of Automated Protection and Monitoring System for Poor Railway Infrastructure" was accepted and presented at "PPG Institute of Technology, Coimbatore, India's" "3rd International Conference on Communication, Computing and Electronics Systems (ICCCES 2021)."Scopus Indexed Series (Springer) "Lecture Notes in Electrical Engineering" will publish the study article. This research is based on developing different sub-models for maximizing railway protection. The main motive of developing the railway protection and monitoring system is to help the railway coordinators and locomotive drivers to ensure proper safety to the railway passengers. Three individual sub-models are accumulated to develop the entire model. Firstly, in order to minimize level crossing accidents, the gateman will be notified about the exact arrival time of the train. In addition, by using sensors, the crossing gate will be automatically closed if any train crosses the level crossing. Moreover, the proposed model can scan the train routes in order to avoid collision accidents with other trains or any obstacles across the route. This entire model used many modern technologies, and it is almost errorless. The countries which possess poor rail infrastructure can implement this cheap and reliable protection and monitoring system in order to avoid major rail accidents. 

This research work was approved and delivered at the University of Georgia's "International Conference on the Fourth Industrial Revolution and Beyond (IC4IR)". The research study will be published in Springer's Scopus-indexed book series. In addition, this conference's research paper was selected as one of the top 100 research papers from 57 nations. This paper is based on the crack and object detection of rail tracks for safe railway transportation in Bangladesh and to reduce the rail accidents due to derailment, which is caused by cracks of the railway tracks. An IoT-based solar-powered automatic crack and object detection vehicle called “Crack and object Identifier Vehicle’’ is proposed to detect the cracks of the rail line and send the coordinates to the station master. As the vehicle examines the rail tracks by image processing method with ESP32 CAM as well as an ultrasonic sensor, IR sensor, and NPN proximity sensors sense to detect the cracks and other components which can be the cause of derailments of the rail line. A dedicated application is built for monitoring the rail tracks and getting the coordinates of the crack location and object detected location to the authority. Also, the coordinates of the location are consigned to the station master by SMS using GSM and GPS modules. In this paper, an automated crack and object identifier vehicle is proposed for detecting cracks and obstacles in the railway track that vehicle can help to reduce the railway accidents from the rail track, especially in those countries whose railway infrastructure is poor.

Ratul, Moloy and Tushar Banik worked together on building a smart firefighting robot with multiple applications. The SAFF robot has dual controlling mechanism. It can operate both manually and autonomously. In manual mode, the robot is manually controlled by a fire fighter to extinguish fire by using water. On the other hand, in the autonomous mode, the robot can automatically detect the fire, walk towards the fire and extinguish it by using carbon di oxide. It is an obstacle avoiding robot which detects fire by using gas and flame sensors. It has two types of fire extinguishers installed inside its body. It can also view what is occurring in front of it through a camera. The entire setup consists of three Arduino boards, buck modules, sensors, a GSM module, relays, servo motors and dc gear motors which all are powered by two 12V lithium polymer batteries.