preventativetestpro-scalable-hybrid-testing-framework-utilizing

PreventativeTestPro: A Scalable Hybrid Testing Framework Utilizing Observability and Generative AI for Proactive Software Quality Engineering

This paper introduces a sophisticated, scalable testing system that integrates observability-driven automation with AI-augmented proactive quality ...

Vishwakarma University

Fast and precise fruit classification or recognition as per quality parameter is the unmet need of agriculture business. This is an open research problem, which always attracts researchers. Machine learning and deep learning techniques have shown very promising results for the classification and object detection problems. Neat and clean dataset is the elementary requirement to build accurate and robust machine learning models for the real-time environment. With this objective we have created an image dataset of Indian fruits with quality parameter which are highly consumed or exported. Accordingly, we have considered six fruits namely apple, banana, guava, lime, orange, and pomegranate to create a dataset. The dataset is divided into three folders (1) Good quality fruits (2) Bad quality fruits, and (3) Mixed quality fruits each consists of six fruits subfolders. Total 19,500+ images in the processed format are available in the dataset. We strongly believe that the proposed dataset is very helpful for training, testing and validation of fruit classification or reorganization machine leaning model.

FruitNet: Indian fruits image dataset with quality for machine learning applications.

Detecting authentic and quality banknotes presents a significant challenge, particularly for individuals with low vision or visual impairments. Extensive research has been dedicated to achieving accurate banknote detection. It is crucial for clean banknotes to be readily detectable and accepted in daily transactions. However, existing Indian currency datasets suffer from limitations, including insufficient size, a lack of datasets on damaged/spoiled banknotes, and the unavailability of publicly accessible datasets featuring spoiled, torn, or altered banknotes. Recognizing the vital importance of a spoiled banknote dataset for the benefit of low vision and visually impaired individuals, we introduce a comprehensive dataset of spoiled banknotes comprising 5125 Indian currency notes. This dataset encompasses both old and new denominations of 10, 20, 50, and 100 Rupees, aiming to significantly enhance the accessibility and accuracy of banknote detection systems. By making this dataset openly accessible to the researchers, we aim to promote research and development of solutions for detection of spoiled banknote.

A comprehensive dataset of damaged banknotes in Indian currency (Rupees) for analysis and classification.

Land Use Land Cover (LULC) classification is pivotal to sustainable environment and natural resource management. It is critical in planning, monitoring, and management programs at various local and national levels. Monitoring changes in LULC patterns over time is crucial for understanding evolving landscapes. Traditionally, LULC classification has been achieved through satellite data by remote sensing, geographic information system (GIS) techniques, machine learning classifiers, and deep learning models. Semantic segmentation, a technique for assigning land cover classes to individual pixels in an image, is commonly employed for LULC mapping. In recent years, the deep learning revolution, particularly Convolutional Neural Networks (CNNs), has reshaped the field of computer vision and LULC classification. Deep architectures have consistently outperformed traditional methods, offering greater accuracy and efficiency. However, the availability of high-quality datasets has been a limiting factor. Bridging the gap between modern computer vision and remote sensing data analysis can revolutionize our understanding of the environment and drive breakthroughs in urban planning and ecosystem change research. The "Sen-2 LULC Dataset" has been created to facilitate this convergence. This dataset comprises of 213,761 pre-processed 10 m resolution images representing seven LULC classes. These classes encompass water bodies, dense forests, sparse forests, barren land, built-up areas, agricultural land, and fallow land. Importantly, each image may contain multiple coexisting land use and land cover classes, mirroring the real-world complexity of landscapes. The dataset is derived from Sentinel-2 satellite imagery sourced from the Copernicus Open Access Hub (https://scihub.copernicus.eu/) platform. It includes spectral bands B4, B3, and B2, corresponding to red, green, and blue (RGB) channels, and offers a spectral resolution of 10 m. The dataset also provides an equal number of mask images. Structured into six folders, the dataset offers training, testing, and validation sets for images and masks. Researchers across various domains can leverage this resource to advance LULC classification in the context of the Indian region. Additionally, it catalyzes fostering collaboration between remote sensing and computer vision communities, enabling novel insights into environmental dynamics and urban planning challenges.

Sen-2 LULC: Land use land cover dataset for deep learning approaches.

The Face Mask Wearing Image Dataset is a comprehensive collection of images aimed at facilitating research in the domain of face mask detection and classification. This dataset consists of 24,916 images, carefully categorized into two main folders: "Correct" and "Incorrect" representing instances of face masks being worn properly and improperly, respectively. Each folder is further divided into four subfolders, each denoting a specific type of face mask - Bandana, Cotton, N95, and Surgical. In the "Correct" folder, images depict individuals correctly wearing their respective face masks, while the "Incorrect" folder contains images of improper face mask usage. To capture variations in face mask application across different demographics, such as age and gender, each subfolder also includes three additional subfolders - Child, Male, and Female. The dataset's diverse content encompasses different face mask types, covering bandana-style, cloth, N95 respirators, and surgical masks, across various age groups and genders. This design ensures a comprehensive representation of real-world scenarios, enabling the evaluation of machine learning algorithms for face mask detection and classification. Researchers can leverage this dataset to develop and assess models that can accurately identify and distinguish between correct and incorrect face mask usage. By contributing to the advancement of face mask detection technologies, this dataset further supports public health initiatives and encourages proper mask-wearing behavior to mitigate the spread of infectious diseases, particularly during times of heightened health concerns such as the COVID-19 pandemic.

Face mask wearing image dataset: A comprehensive benchmark for image-based face mask detection models.

The United Nations advocates for sustainable urban planning and design, emphasizing green infrastructure initiatives to mitigate urban heat island effects and enhance the resilience and livability of cities globally. To address urban heat challenges, a study was conducted in Chennai, India, from April to June 2023. The study focused on assessing temperature dynamics on a building's terrace by comparing a well-maintained garden area with an exposed region. Temperature and humidity sensors were deployed in both the garden and exposed areas of the terrace, as well as within rooms beneath it, to monitor hourly temperature fluctuations. The findings indicate a significant reduction in internal room temperatures in areas with rooftop gardens, ranging from 4 to 11 °C, depending on the time of year and sun's position, compared to rooms with fully exposed roof configurations. Additionally, simulation studies were performed to validate these findings, suggesting that optimizing the distribution of soil beds and plant density across the roof could yield an additional temperature reduction of 3-4 °C, resulting in an overall difference of up to 14-15 °C. The study highlights the efficacy of rooftop gardens in providing cooling effects during daylight hours and maintaining temperature parity post-sunset. Through analysis of sensor data, the research elucidates the intricate relationship between green infrastructure and thermal comfort, offering insights for energy-efficient building design and resilient urban planning. The findings underscore the potential of rooftop gardens in fostering a more comfortable, energy-efficient, and sustainable urban living environment.

Mitigating urban heat island and enhancing indoor thermal comfort using terrace garden.

Fake-news detection requires systems that are multilingual, multimodal, and explainable-yet the majority of the existing models are English-centric, text-only, and opaque. This study introduces two key innovations: (i) a new multilingual-multimodal dataset of 74,000 news articles in Hindi, Gujarati, Marathi, Telugu, and English with paired images, and (ii) Hybrid Explainable Multimodal Transformer Fake (HEMT-Fake) that integrates text, image, and relational signals with hierarchical explainability. The architecture combines transformer embeddings, a convolutional neural network-bidirectional long short-term memory (CNN-BiLSTM) text encoder, residual network (ResNet) image features, and graph sample and aggregate (GraphSAGE) metadata, all of which are fused via multi-head attention. Its explainability module unites attention, Shapley Additive exPlanations (SHAP), and local interpretable model-agnostic explanations (LIME) to provide token-, sentence-, and modality-level transparency. Across four languages, HEMT-Fake delivers a ~ 5% Macro-F1 improvement over Cross-Lingual Language Model with RoBERTa (XLM-R) architecture and Multilingual Bidirectional Encoder Representations From Transformers (mBERT), with gains of 7-8% in low-resource languages. The model achieves 85% accuracy under adversarial paraphrasing and 80% on artificial intelligence (AI)-generated fake news, halving robustness losses compared to baselines. Human evaluation reveals that 82% of explanations are judged to be meaningful, confirming transparency and trust for fact-checkers.

Explainable multilingual and multimodal fake-news detection: toward robust and trustworthy AI for combating misinformation.

Ensuring the post-harvest quality and health of leafy vegetables is critical for minimizing economic loss, enhancing food security, and promoting sustainable agricultural practices. Spinach, a highly nutritious yet perishable crop, is particularly susceptible to rapid freshness degradation and foliar diseases. While computer vision and deep learning have shown promise for automated quality assessment, existing models often lack the robustness to handle the dual-task classification of both freshness and disease states across diverse local spinach varieties. To bridge this gap, this paper introduces a novel hybrid Convolutional Neural Network (CNN) architecture specifically designed for the multi-class detection of freshness and visual disease symptoms in local spinach leaves. The proposed model synergistically integrates a powerful feature extraction backbone with a tailored attention and fusion mechanism, enhancing its ability to capture discriminative spatial and textural features critical for fine-grained classification. It was trained and validated on a curated dataset comprising high-resolution images of three prominent local varieties (Malabar, Water, and Red spinach) in both fresh and non-fresh conditions. The proposed hybrid model achieved a classification accuracy of 98.36%, significantly outperforming benchmark state-of-the-art models including DenseNet121, ResNet50, and EfficientNetB0. Furthermore, explainable AI (XAI) techniques visually validated the model’s decision-making process, confirming its focus on biologically relevant leaf regions. The results demonstrate that the proposed hybrid framework offers a highly accurate, reliable, and interpretable tool for non-destructive, real-time quality monitoring. This work provides a significant contribution towards intelligent post-harvest management systems, capable of reducing waste and supporting the value chain for local spinach cultivation.

Article	Year
FruitNet: Indian fruits image dataset with quality for machine learning applications. Data in brief\| PubMed ID: 34917715	2022
A comprehensive dataset of damaged banknotes in Indian currency (Rupees) for analysis and classification. Data in brief\| PubMed ID: 37920385	2023
Sen-2 LULC: Land use land cover dataset for deep learning approaches. Data in brief\| PubMed ID: 37965594	2023
Face mask wearing image dataset: A comprehensive benchmark for image-based face mask detection models. Data in brief\| PubMed ID: 38075619	2023
Mitigating urban heat island and enhancing indoor thermal comfort using terrace garden. Scientific reports\| PubMed ID: 38678098	2024
Explainable multilingual and multimodal fake-news detection: toward robust and trustworthy AI for combating misinformation. Frontiers in artificial intelligence\| PubMed ID: 41451323	2025
A hybrid CNN model for multi-class freshness and disease detection in local spinach varieties. BMC plant biology\| PubMed ID: 41663956	2026

Vishal Meshram

Biography

JoVE Journal Publications

Other Publications