Journal of Scientific Research and Technology

Application of Deep Learning Technique for Tomato Maturity Stage Prediction

Ankit — 2025-07-01

Tomatoes are a major crop worldwide, and accurately classifying their maturity is important for many agricultural
applications, such as harvesting, grading, and quality control. Primary objective to develop an automated system
capable of identifying various defects in tomatoes and providing relevant treatment suggestions. Leveraging deep
learning techniques, a CNN model is trained to classify tomatoes into four categories: Damaged, Old, Ripen, and
Unripen. The implementation involves training a convolutional neural network and testing dataset of tomato images,
for classification, and deploying model intended real-time predictions. Project has potential to improve effectiveness of
tomato harvesting & reduce waste. The CNN model then predicts the stage of the tomato, providing the probability of
the prediction. Alongside the classification result, the system offers detailed information on the cause of the defect,
appropriate treatment methods, and nutritional content, aiding users in making informed decisions regarding the
tomatoes usability. The implementation ensures a robust and efficient detection mechanism, even in varying lighting
conditions and backgrounds. An Accuracy of 94.48% was attained when taught sculpt was used to make prediction on
the test dataset.

Multifunctional Agricultural Robot

Neelambika — 2025-07-01

This paper presents the design, development, and implementation of a multipurpose agricultural robot aimed at
automating essential farming tasks to enhance productivity and reduce labor. The robot is capable of digging soil,
levelling mud, and spraying water and fertilizers, operating on a combination of battery and solar energy. Key
components include a relay switch, Bluetooth modules for user interaction, and various sensors to ensure precise
operation. The hands-free, efficient design addresses the increasing need for innovative solutions in agriculture, driven
by the challenges of labor intensive traditional farming. By leveraging advanced technology, the proposed system
demonstrates significant potential in improving farming efficiency and sustainability, particularly in the context of the
growing interest in autonomous agricultural vehicles. The robot's hands-free, autonomous capabilities make it a
valuable asset for farmers, reducing the physical effort and time required for traditional farming practices.

A Novel Approach for Real-Time Audio-to-Sign Language Translation using Naive Bayes classifier and Natural Language Processing Technique

Poornima — 2025-07-01

Real-time translation from spoken language to sign language, crucial for the deaf and hard of hearing, is difficult employing machine learning advances, this research suggests employing Naive Bayes to instantaneously translate audio input into sign language motions. The initiative provides precise translations between spoken and visual language to improve accessibility and inclusion. The system generates sign language animations from audio inputs using natural language processing and deep learning. Language extraction, tense and context classification, and proficient sign language synthesis are key aspects. The research innovates by using a Naive Bayes model trained on varied spoken language and sign language gesture datasets to ensure robust performance across contexts and user situations. To enhance realism and comprehension, the system integrates a 3D animated avatar to render the translated signs visually. A custom-built gesture mapping mechanism aligns each keyword with its corresponding sign language motion. The tool operates in real-time, supporting both sentence-level and word-level translations. This approach contributes to bridging communication gaps and supports inclusive digital interaction for the hearing-impaired community.

Automated Detection of Image Forgery with Deep Learning

Shreedevi S Patil — 2025-07-01

The image forgery techniques are used to provide the particular image in the form of computerized pictures with no errors in capturing of digital image information which shows the original image. While compared to normal images the edited images are difficult to find out the forged images. The image should maintain the authenticity and integrity to secure the picture from unauthorized users. In this paper, we have compared the digital image forgery and JPEG is the most common format used by the photographic images and the digital camera devices. These operations are performed in an adobe photo-shop using with the content of image security to restore some digital image with an authenticity and integrity to detect the digital image forgery using active and passive techniques.

Smart Traffic Control Approach For Emergency Vehicles

Dr Rashmi Patil — 2025-07-01

Traffic congestion problem is a phenomenon which contributed huge impact to the transportation system in our country. This causes many problems especially when there are emergency cases at traffic light intersections which are always busy with many vehicles. A traffic light assistance system is designed in order to solve these problems. This system was designed to be operated when it received signal from emergency vehicles based on radio frequency (RF) transmission and used the Programmable Arduino Atmega 328 microcontroller to controls the LEDs used in the traffic signals. The use of hazard LED in the system which helps the emergency vehicles to pass the traffic easily. This system will reduce accidents which often happen at the traffic light intersections because of another vehicle had to huddle for given a special route to emergency vehicle. As the result, this project successful analyzing and implementing the traffic assistance system for emergency vehicles.

CNN-Based Tomato Leaf Disease Detection With Smart Spraying Mechanism

Swati V Partapur — 2025-07-01

Indian agriculture plays a vital role in crop production and economic development, providing employment to a significant portion of the population. The health of plants is crucial for better yield and profit, requiring continuous monitoring to prevent diseases that can severely impact crop productivity. Traditional disease detection relies on manual observation, which is time-consuming and inefficient. To address this, an automated plant disease identification and prevention system is introduced, integrating IoT-based agricultural monitoring with Machine Learning algorithms. Various IoT sensors, such as thermal, moisture, humidity, and color sensors, help detect plant diseases at an early stage. The system enables timely intervention through automated medicine spraying, improving efficiency and crop health.

Advanced Structural Analysis And Design Optimization Of High- Rise Building Utilizing Building Information Modeling (BIM) Techniques

Ritik Yadav — 2025-07-01

In this study, we will be exploring the advanced structural analysis and design optimization techniques for a G+16 high-rise RCC building using STAAD.Pro V8i SS6. Based on IS codes, the building is evaluated for dead, live, and seismic loads, with manual calculations and software evaluations being utilized to evaluate the data. This involves evaluating the design elements, including stresses on beams and columns, reinforcement detailing, and structural stability. Other important considerations are also given. These findings provide evidence that STAAD.Pro can design high-rise buildings with accuracy, safety, and affordability without requiring manual methods, while also being an excellent substitute for advanced software tools.

Studying The Strength And Durability Properties Of Steel Fiber Reinforced Self -Compacting Concrete

Agrim — 2025-07-01

This study investigates the strength and durability characteristics of steel fiber reinforced self-compacting concrete (SFRSCC). The primary objective is to evaluate the influence of varying steel fiber contents on the mechanical performance and durability performance of self-compacting concrete. Tests conducted include compressive strength, splitting tensile strength, flexural strength, water absorption, sorptivity, and resistance to chloride ion penetration. The experimental results demonstrate that the inclusion of steel fibers significantly enhances the tensile and flexural strengths, improves the concrete's crack resistance, and reduces permeability, thereby augmenting its durability. The findings suggest that steel fiber reinforcement is an effective method to improve the structural integrity and longevity of self-compacting concrete in various construction applications. The optimal fiber content balancing workability, strength, and durability is also discussed, providing valuable insights for structural engineers and material scientists aiming to design durable, high-performance concrete composites.

Research On The Development Of Eco-Friendly, High-Strength, And Durable Concrete Materials Used In Modern Construction

Devdhar Dwivedi — 2025-07-01

The construction industry is increasingly seeking sustainable solutions to address environmental concerns while meeting the demands for high-performance structural materials. Traditional concrete, although widely used, significantly contributes to carbon emissions and resource depletion. This research explores the development of eco-friendly concrete composites that combine high strength and durability for modern construction applications. Emphasis is placed on incorporating supplementary cementitious materials such as fly ash, slag, and silica fume, as well as innovative eco-friendly binders and admixtures that reduce the reliance on Portland cement. The study evaluates the mechanical properties, durability factors, and environmental impacts through experimental testing, including compressive strength, tensile strength, water absorption, and resistance to chemical attacks. Sustainable approaches such as the utilization of recycled aggregates and biogenic materials are also examined. The findings demonstrate that optimized mixes can achieve comparable or superior performance to conventional concrete, with significant reductions in carbon footprint and resource consumption. The research underscores the potential of eco-friendly, high-strength, and durable concrete as a cornerstone for sustainable infrastructure development, contributing to environmentally responsible construction practices.

Agricultural Pesticides Spraying Remote Control Robot

Asst. Prof. Nandini — 2025-07-02

Automatic pesticide sprayers offer several significant advantages in modern agriculture. They enhance efficiency by covering larger areas quickly and reducing labor costs. These machines also improve precision in pesticide application, minimizing wastage and environmental impact. Furthermore, they enhance safety foe farmers by reducing exposure to harmful chemicals. By optimizing resource use and minimizing risks, automatic sprayers contribute to sustainable and productive agricultural practices. In conclusion, agriculture pesticides spraying smart remote-control robots offer a range of benefits, including increased efficiency, improved safety, enhanced accuracy, cost-effectiveness, and environmental benefits. These robots can help farmers optimize pesticide application, reduce waste, and promote healthier crops. As technology continues to evolve, we can expect to see further advancements in this field, leading to more sustainable and productive agriculture practices.

A Machine Learning Approach To Classify Medicinal Plant Leaf By Using Random Forest And KNN

Bhagyashree — 2025-07-02

Numerous plant and herb types partake remained cast-off broadly in traditional medicine due to their medicinal properties and cost-effectiveness compared to modern prescription drugs. Nevertheless, accurately identifying medicinal plants poses a significant challenge in the realm of machine learning. This project leverages a comprehensive dataset of leaf images from various medicinal plants to train and evaluate a classification model.The primary impartial of this explore is to develop an advanced ML model capable of accurately identifying different Indian medicinal plants. By focusing on several vulnerable species, the study employs the Random Forest and K-Nearest Neighbors algorithms to classify the plants. The KNN algorithm achieved an accuracy of 84.33%, while the Random Forest algorithm outperformed it with an accuracy of 85.82%, demonstrating its superior effectiveness.Beyond mere identification, the project explores the medicinal properties & aids related with individually recognized plant. Given the visual similarities among many plant species, classification is predominantly stimulating, expressly when commerce with extensive datasets. This research aims to bridge the gap between traditional botanical knowledge and modern technological advancements, thereby contributing to the preservation and dissemination of valuable medicinal information.By utilizing a Random Forest-based predictive model, this structure not only aids in the accurate identification of medicinal plants but also supports the conservation of ancient and traditional medicinal knowledge. This integration of historical wisdom with contemporary machine learning techniques ensures that valuable botanical information is preserved and made accessible for future generations

An Intelligent Video Analysis Framework For Classifying And Prioritizing Harmful Social Media Content With CNNs

Soumya — 2025-07-02

Social media platform have develop essential channels intended communication, content sharing, and community building. However, the widespread distribution of user-generated content introduces significant challenges in monitoring and managing harmful, inappropriate, or violative material. Addressing these challenges is crucial to maintaining a safe and respectful online environment. This project presents a comprehensive framework for detecting and rating violative user content in social media, employing advanced computer vision techniques to analyze video content. The framework extracts frames from videos and utilizes (CNNs) to classify various forms of violations with high accurateness. A robust dataset, encompassing diverse categories of violations, is employed to train the model, ensuring its effectiveness across different contexts and platforms. The framework's evaluation component is thorough, incorporating system of measurement such as accurateness, precision, recall, & F1-score to assess model performance. Confusion matrices & classification reports offer detailed insights into the system's effectiveness. The model's capability to process video frames in real- time simplifies its integration into prevailing social media monitoring systems, providing a scalable solution for content moderation. In addition to detection, the framework includes a rating mechanism that evaluates the severity of detected violations. This rating system aids in prioritizing content review processes, ensuring that the most harmful material is addressed promptly. The use of advanced machine learning algorithms and comprehensive training data allows the framework to adapt to evolving content trends and emerging threats effectively. Overall, this project delivers a scalable, accurate, and efficient solution for detecting and rating violative user content on social media platforms. By enhancing content moderation capabilities, it contributes significantly to the creation of safer and more respectful online communities.

Advanced CNN Approach To Automated Squint Eye Detection And Comprehensive Early Intervention

Prakruti — 2025-07-02

Squint eye, also recognized as strabismus, is condition in which eyes do not properly align by each other while looking at an object. This misalignment can result in both eyes pointing in different directions, leading to double vision or a loss of depth perception. Early detection & treatment of squint eye are crucial to prevent long-term visual impairments & improve superiority of life intended affected individuals. Traditional methods for diagnosing squint eye involve subjective assessments by medical professionals, which can be time-consuming & prone to variability. To address these challenges, a novel approach using CNN intended automated squint eye detection has been developed. This study presents a comprehensive framework for detecting squint eye using CNNs, leveraging the powerful capabilities of deep learning to analyze and classify eye images with high accuracy. The proposed system involves the collection and preprocessing of bulky dataset of eye metaphors, which be subsequently worn to train a CNN model. Model is intended to identify key features indicative of squint eye, such as positioning of pupils & alignment of eyes. Once trained, replica is capable of analyzing new eye images to establish presence of squint eye with high precision. Performance of developed CNN model is evaluated using variety of metrics, including accurateness, sympathy, specificity, & F1-score. Results exhibit that representation achieves noteworthy improvement over traditional diagnostic methods, providing a reliable and efficient tool for squint eye detection. Additionally, the study explores the potential applications of this technology in clinical settings, highlighting its benefits for early diagnosis and intervention. Regarding licensing regulations, it imperative to note that individuals with squint eye may face restrictions when it comes to obtaining a driving license. In many regions, the Road Transport Office (RTO) does not issue driving licenses to individuals with squint eye due to prospective risks associated with impaired vision, such as double vision and reduced depth perception. Use of automated detection systems can aid in early diagnosis & appropriate management of this condition, ensure that pretentious individuals receive necessary treatment and support.

AI KNOWLEDGE HUB 2.0

Swati S Kagi — 2025-07-02

The AI Knowledge Hub is an interactive application that enables users to access and analyze multimedia content like PDFs, YouTube videos, and images. It uses advanced AI models and natural language processing to extract relevant information from these sources. The system extracts text from PDFs using PyPDF2, retrieves video transcripts via the YouTube Transcript API, and processes images using Optical Character Recognition (OCR). The extracted text is processed using LangChain and stored in FAISS, a vector database optimized for fast similarity searches. The Gemini AI model analyzes user queries, generating context-aware responses. The application supports multilingual translation using Googletrans and a text-to-speech (TTS) feature powered by gTTS. The intuitive Streamlit-based interface allows for quick navigation. The AI Knowledge Hub is designed for real-time processing, making it useful for students, researchers, and professionals. Future enhancements include support for additional file formats, cloud storage integration, advanced video analysis, offline processing capabilities, and compatibility with voice assistants like Alexa and Google Assistant.