intelligent-federated-learning-framework-for-non-colocated

Intelligent Federated Learning Framework for Non-colocated and Heterogeneous Datasets

Federated learning has significant potential for distributed model training while preserving privacy, but it faces challenges related to convergence, ...

MIT Art Design and Technology University

Indian bay leaf is a crucial spice in Indian cuisine. However, its quality and authenticity are often compromised. To address this, we introduce The Digital Indian Bay leaf dataset, a comprehensive collection of high-resolution 5696 images capturing diverse bay leaf samples under controlled conditions. The dataset encompasses variations in leaf conditions such as fresh leaf, dried leaf and diseased prone leaf. The dataset is meticulously curated to support research in condition analysis and machine learning applications for leaf quality assessment. To ensure data diversity, each category includes a wide range of images captured under controlled conditions with varying lighting, background, and leaf orientation. By providing a standardized and accessible resource, this dataset aims to accelerate research in this domain and contribute to the improvement of the Indian spice industry.

A dataset revolutionizing Indian bay leaf analysis.

 is a significant medicinal herb extensively used in traditional oriental medicine and gaining global popularity. The primary constituents of  leaves are triterpenoid saponins, which are predominantly believed to be responsible for its therapeutic properties. Ensuring the use of high-quality leaves in herbal medicine preparation is crucial across all medicinal practices. To address this quality control issue using machine learning applications, we have developed an image dataset of  leaves. The images were captured using Samsung Galaxy M21 mobile phones and depict the leaves in "Dried," "Healthy," and "Unhealthy" states. These states are further divided into "Single" and "Multiple" leaves categories, with "Single" leaves being further classified into "Front" and "Back" views to facilitate a comprehensive study. The images were pre-processed and standardized to 1024 × 768 dimensions, resulting in a dataset comprising a total of 9094 images. This dataset is instrumental in the development and evaluation of image recognition algorithms, serving as a foundational resource for computer vision research. Moreover, it provides a valuable platform for testing and validating algorithms in areas such as image categorization and object detection. For researchers exploring the medicinal potential of  in traditional medicine, this dataset offers critical information on the plant's health, thereby advancing research in herbal medicine and ethnopharmacology.

Dataset of  leaves for quality assessment and machine learning applications.

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.

Although artificial intelligence-based cancer diagnostic models have demonstrated strong predictive performance, their lack of transparency and reliance on single-modality data continue to limit clinical trust and adoption. Effectively integrating multi-modal data with interpret-able decision-making remains a key challenge.

Article	Year
A dataset revolutionizing Indian bay leaf analysis. Data in brief\| PubMed ID: 39554547	2024
Dataset of leaves for quality assessment and machine learning applications. Data in brief\| PubMed ID: 39687371	2024
Explainable multilingual and multimodal fake-news detection: toward robust and trustworthy AI for combating misinformation. Frontiers in artificial intelligence\| PubMed ID: 41451323	2025
Explainable multi-modal deep learning for transparent cancer diagnosis: integrating radiology, clinical features, and decision visualization. Frontiers in artificial intelligence\| PubMed ID: 41809581	2026

Amol A. Bhosle

Biography

JoVE Journal Publications

Other Publications