Leveraging knowledge distillation for lightweight and interpretable deep learning in Ethiopian medicinal plant classification

Medicinal Plant species help to cure various diseases across the world. The automated identification of medicinal plant species to treat disease based on their structure is much required in pharmaceutical laboratories. Plant Species with a complex background in the field will make the detection and classification more difficult. In this paper, optimization of bacterial foraging technique has been employed towards medicinal plant prediction and classification architecture based on feed-forward neural network. It is capable of identifying both complex structures of medicinal plants. Feed-forward Neural Networks are considered to have good recognition accuracy compared to other machine learning approaches. Further bacterial foraging has been implemented to minimize the feature search space to the classifier and provides optimal features for the plant classification. The experimental outcomes of the proposed approach has been analysed by employing the medley dataset and evaluating the performance of the proposed approach with respect to dice similarity coefficient, Specificity and sensitivity towards medicinal plant classification. The findings are very positive, and further research will focus on using a large dataset and increased computing resources to examine how well deep-learning neural networks function in identifying medicinal plants for use in health care.

There is a global dependence on medicinal plants for the treatment of diverse health conditions, particularly in developing countries, where government medical facilities are inadequate. Conventional machine learning techniques and traditional methodologies are inadequate for rapid, accurate, and reliable identification and classification of medicinal plants. This has paved the way for deep learning(DL) models to identify and classify medicinal plants. This review focuses on the following questions: 1) Which deep learning models were used for medicinal plant identification and classification and how did they perform? 2) Did hybrid models perform better than ensemble models? 3) What common datasets are used for deep learning-based medicinal plant identification and classification studies? 4) Which data augmentation techniques were prevalent in medicinal plant identification and classification studies? 5) Which performance metrics are frequently used to evaluate medicinal plant identification models? 6) What data splitting ratios are commonly used? 7) Was transfer learning prevalent in medicinal plant identification and classification studies? This review complements the limited literature on medicinal plant identification and classification using deep learning. This study analyzes gaps in the existing literature and provides recommendations for future research. This will enhance the development of robust deep learning models to identify and classify medicinal plants for indigenous knowledge preservation, biodiversity monitoring, and their proper use in pharmaceutical processes.

Medicinal plants are the backbone of the system of medicines; they are the richest bioresource of drugs of traditional systems of medicine, modern medicines, nutraceuticals, food supplements, folk medicines, pharmaceutical intermediates, and chemical entities for synthetic drugs. These plants are classified according to their medicinal values. Classification of medicinal plants is acknowledged as a significant activity in the production of medicines along with the knowledge of its use in the medicinal industry. Medicinal plant classification based on parts such as leaves has shown significant results. An automated system for the identification of medicinal plants from leaves using Image processing and Machine Learning techniques has been presented. This paper provides knowledge of the process of identification of medicinal plants from features extracted from the images of leaves and different preprocessing techniques used for feature extraction from a leaf. Many features were extracted from each leaf such as its length, width, perimeter, area, color, rectangularity, and circularity. It is expected that for the automatic identification of medicinal plants, a web-based or mobile computer system will help the community people to develop their knowledge on medicinal plants, help taxonomists to develop more efficient species identification techniques and also participate significantly in the pharmaceutical drug manufacturing.

The accurate identification of medicinal plants is crucial for ensuring the quality and efficacy of herbal remedies. This paper investigates the application of deep learning for automatic medicinal plant classification using leaf images. We propose a deep learning model based on the pre-trained ResNet-50 architecture to classify medicinal plants from the LeafSnap dataset. The model leverages transfer learning to exploit pre-trained features and fine-tune them for the specific task of medicinal plant identification. We evaluate the performance of the proposed model and achieve a high accuracy of approximately 99.86%. This demonstrates the effectiveness of deep learning, particularly the ResNet-50 architecture, for automated medicinal plant classification. Our findings highlight the potential of this approach for applications such as supporting field identification, streamlining herbarium workflows, and potentially aiding in the development of novel drug discovery pipelines. Keywords- Medicinal plants, Deep learning, Convolutional Neural Networks (CNNs), ResNet-50, Transfer learning, Data augmentation, LeafSnap dataset, Image classification

Interpretable deep learning (DL) models that can provide biological insights, in addition to accurate predictions, are of great interest to the biomedical community. Recently, interpretable DL models that incorporate signaling pathways have been proposed for drug response prediction (DRP). While these models improve interpretability, it is unclear whether this comes at the cost of less accurate DRPs, or a prediction improvement can also be obtained. We comprehensively and systematically assessed four state-of-the-art interpretable DL models using three pathway collections to assess their ability in making accurate predictions on unseen samples from the same dataset, as well as their generalizability to an independent dataset. Our results showed that models that explicitly incorporate pathway information in the form of a latent layer perform worse compared to models that incorporate this information implicitly. However, in most evaluation setups, the best performance was achieved using a black-box multilayer perceptron, and the performance of a random forests baseline was comparable to those of the interpretable models. Replacing the signaling pathways with randomly generated pathways showed a comparable performance for the majority of the models. Finally, the performance of all models deteriorated when applied to an independent dataset. These results highlight the importance of systematic evaluation of newly proposed models using carefully selected baselines. We provide different evaluation setups and baseline models that can be used to achieve this goal. Implemented models and datasets are provided at https://doi.org/10.5281/zenodo.7787178 and https://doi.org/10.5281/zenodo.7101665, respectively.

Plants are the basis of all living things on earth, supplying us with oxygen, food, shelter, medicine, and preserving the planet from dam-ages that could face climate changes. Concerning their medicinal abilities, limited access to proper medical centers in many rural areas and developing countries made traditional medicine preferable by the community. In addition, their lower side effect and affordability also plays a big role. More than half of the population uses medicinal plants directly and indirectly for animals and personal use in Ethiopia. However, accurate medicinal plant identification has always been a challenge for manual identification and automatic recognition systems mainly because the knowledge transfer between the knowledge holders (traditional physicians, elderly) and modern science have a huge gap. Several studies addressed an automatic plant recognition system using different feature extraction methods and classification algorithms. In this paper, a novel dataset, which was based on Ethiopian medicinal plants, that use the leaf part of the plant, as a medicine was used to automatically classify the plants accordingly using their leaf image. An attempt has been made to collect leaf images of medicinal plants in Ethiopia, to train, test collected dataset images, and classify those images using convolutional neural network models like GoogleNet and AlexNet. The proposed convolutional neural networks were fine-tuned with the adjustment of hyper-parameters like learning rate, the number of epochs, optimizers to the models. Image augmentation is also implemented to enlarge the dataset. The experimental result for the augmented dataset and more training epoch gave better performance and accuracy in the classification of the images. From the two selected convolutional neural network models, the best model is then determined based on the result in accuracy and loss; from an experiment conducted, the best model, which is GoogLeNet with an accuracy of 96.7 % chosen to develop a web-based automatic medicinal plant classification system.KeywordsMedicinal plantsLeaf imagesConvolutional neural networkGoogleNetAlexNet

Identifying the correct medicinal plants that goes into the preparation of a medicine is very important in the ayurvedic, folk, and herbal medicinal industry. Botanists invest a lot of time in identifying plant species by direct observation. Recognition of medicinal plants among various plant species is very difficult for ordinary people. So, in order to overcome these difficulties, we are developing an AI-based Automatic Classification system for classifying medicinal plants among the several plants. This is mostly useful for the society to identify the ayurvedic leaves, which can be used in traditional medicine. By using this system, normal people can easily recognize these medicinal plants. This study outlines a technique for classifying different medicinal plant species using color images of some medicinal plant species. With the aid of the pre-trained classifier VGG-19, the task is carried out utilizing transfer learning to increase accuracy. Image pre-processing, image augmentation, feature extraction, and recognition are the four main classification steps that are carried out as part of the overall model evaluation. By using pre-defined hidden layers like convolutional layers, max-pooling layers, and fully connected layers, the VGG-19 classifier is able to understand the features of leaves. After that, the soft-max layer is used to create a feature representation for all plant classes. In order to help estimate the correct class of an unidentified medicinal plant, the model gathers information about various medicinal plants, which contains around nineteen different classes. This system will classify the medicinal plant species with high accuracy. Identification and classification of medicinal plants are essential for better treatment.

Accurate classification of medicinal plant images into high-level categories and specific sub-groups is essential for various applications, including agriculture, plant research, and conservation. This paper proposes a multi-stage deep learning approach to enhance the precision of medicinal plant image classification. In the first stage, known as Broad Classification, CNN and pre-trained models such as VGG16, ResNet50 and EfficientNetB0 are utilized to categorize images into high-level groups, including "Medicinal Plants," "Fruit-Related Plants," and "Flower-Related Plants." The model is fine-tuned using data augmentation techniques to ensure robust learning and generalization. In the second stage, referred to as Detailed Classification, separate models are trained for each high-level group to classify images into specific sub-groups within that category. The architecture of these models is adjusted to accommodate the unique number of classes in each sub-group. Each model undergoes training with optimized hyperparameters and is evaluated based on precision, recall, F1-score, and accuracy. The proposed multi-stage method demonstrates the ability to handle both broad and fine-grained medicinal plant classifications effectively, showcasing an improvement in classification performance over traditional single-stage models. This approach highlights the potential for deep learning to contribute to more precise and practical medicinal plant image classification solutions.

Ayurvedic medicines have a vital role in preserving physical and mental health of human beings. Identification and classification of medicinal plants are essential for better treatment. Lack of experts in this field makes proper identification and classification of medicinal plants a tedious task. Hence, a fully automated system for medicinal plant classification is highly desirable. This work proposes AyurLeaf, a Deep Learning based Convolutional Neural Network (CNN) model, to classify medicinal plants using leaf features such as shape, size, color, texture etc. This research work also proposes a standard dataset for medicinal plants, commonly seen in various regions of Kerala, the state on southwestern coast of India. The proposed dataset contains leaf samples from 40 medicinal plants. A deep neural network inspired from Alexnet is utilised for the efficient feature extraction from the dataset. Finally, the classification is performed using Softmax and SVM classifiers. Our model, upon 5-cross validation, achieved a classification accuracy of 96.76% on AyurLeaf dataset. AyurLeaf helps us to preserve the traditional medicinal knowledge carried by our ancestors and provides an easy way to identify and classify medicinal plants.

This study presents a new approach to understand the causes of groundwater drought events with interpretable deep learning (DL) models. As prerequisites, accurate long short‐term memory (LSTM) models for simulating groundwater are built for 16 regions representing three types of spatial scales in the southeastern United States, and standardized groundwater index is applied to identify 233 groundwater drought events. Two interpretation methods, expected gradients (EG) and additive decomposition (AD), are adopted to decipher the DL‐captured patterns and inner workings of LSTM networks. The EG results show that: (a) temperature‐related features were the primary drivers of large‐scale groundwater droughts, with their importance increasing from 56.1% to 63.1% as the drought events approached from 6 months to 15 days. Conversely, precipitation‐related features were found to be the dominant factors in the formation of groundwater drought in small‐scale catchments, with the overall importance ranging from 59.8% to 53.3%; (b) Seasonal variations in the importance of temperature‐related factors are inversely related between large and small spatial scales, being more significant in summer for larger regions and in winter for catchments; and (c) temperature‐related factors exhibited an overall “trigger effect” on causing groundwater drought events in the studying areas. The AD method unveiled how the LSTM network behaved differently in retaining and discarding information when emulating different groundwater droughts. In summary, this study provides a new perspective for the causes of groundwater drought events and highlights the potential and prospect of interpretable DL in enhancing our understanding of hydrological processes.

This study aims to develop and validate an interpretable deep learning (DL) model and a nomogram based on endoscopic ultrasound (EUS) images for the prediction of pathological grading in pancreatic neuroendocrine tumors (PNETs). This multicenter retrospective study included 108 patients with PNETs, who were divided into train (n = 81, internal center) and test cohorts (n = 27, external centers). Univariate and multivariate logistic regression were used for screening demographic characteristics and EUS semantic features. Deep transfer learning was employed using a pre-trained ResNet18 model to extract features from EUS images. Feature selection was conducted using the least absolute shrinkage and selection operator (LASSO), and various machine learning algorithms were utilized to construct DL models. The optimal model was then integrated with clinical features to develop a nomogram. The performance of the model was assessed using the area under the curve (AUC), calibration curves, decision curve analysis (DCA), and clinical impact curves (CIC). The nomogram, which integrates the optimal DL model (Naive Bayes) with clinical features, achieved AUC values of 0.928 (95% CI 0.849–0.981) in the train cohort and 0.882 (95% CI 0.778–0.954) in the test cohort. Calibration curves revealed minimal discrepancies between predicted and actual probabilities, with mean absolute errors of 4.5% and 6.6% in the train and test cohorts, respectively. DCA and CIC demonstrated substantial net benefit and clinical utility. The SHapley Additive exPlanations (SHAP) method provided insights into the contribution of each DL feature to the model’s predictions. This study developed and validated a novel interpretable DL model and nomogram using EUS images and machine learning, which holds promise for enhancing the clinical application of EUS in identifying PNETs’ pathological grading.

Medicinal Plants Identification Using Federated Deep Learning

The identification and classification of medicinal plants are crucial for botanical research, traditional medicine, and AI-driven applications. However, the absence of a standardized, high-quality dataset limits advancements in automated species recognition. This study introduces SIMPD Version 1 (South Indian Medicinal Plants Dataset), a curated dataset comprising high-resolution images of diverse medicinal plant species native to South India. The dataset integrates detailed taxonomic classifications and metadata to facilitate precise species identification and biodiversity analysis. Images were acquired under real-world conditions, considering variations in illumination, pose, and environmental factors to enhance dataset robustness. SIMPD is designed to support machine learning applications, particularly in image-based plant classification, object detection, and segmentation tasks. By providing an extensive dataset for AI-driven research, this work aims to bridge the gap between traditional ethnobotanical knowledge and modern computational methodologies, fostering advancements in medicinal plant classification, conservation, and ecological research.

The identification of Chinese medicinal plants was conducted to rely on ampelographic manual assessment by experts. More recently, machine learning algorithms for pattern recognition have been successfully applied to leaf recognition in other plant species. These new tools make the classification of Chinese medicinal plants easier, more efficient and cost effective. This study showed comparative results between machine learning models obtained from two methods: i) a morpho-colorimetric method and ii) a visible (VIS)/Near Infrared (NIR) spectral analysis from sampled leaves of 20 different Chinese medicinal plants. Specifically, the automated image analysis and VIS/NIR spectral based parameters obtained from leaves were used separately as inputs to construct customized artificial neural network (ANN) models. Results showed that the ANN model developed using the morpho-colorimetric parameters as inputs (Model A) had an accuracy of 98.3% in the classification of leaves for the 20 medicinal plants studied. In the case of the model based on spectral data from leaves (Model B), the ANN model obtained using the averaged VIS/NIR spectra per leaf as inputs showed 92.5% accuracy for the classification of all medicinal plants used. Model A has the advantage of being cost effective, requiring only a normal document scanner as measuring instrument. This method can be adapted for non-destructive assessment of leaves in-situ by using portable wireless scanners. Model B combines the fast, non-destructive advantages of VIS/NIR spectroscopy, which can be used for rapid and non-invasive identification of Chinese medicinal plants and other applications by analyzing specific light spectra overtones from leaves to assess concentration of pigments such as chlorophyll, anthocyanins and others that are related active compounds from the medicinal plants. Keywords: ampelography, computer vision, artificial neural networks, pattern recognition, Chinese medicinal plants DOI: 10.25165/j.ijabe.20191202.4637 Citation: Xue J R, Fuentes S, Poblete-Echeverria C, Viejo C G, Tongson E, Du H J, et al. Automated Chinese medicinal plants classification based on machine learning using leaf morpho-colorimetry, fractal dimension and visible/near infrared spectroscopy. Int J Agric & Biol Eng, 2019; 12(2): 123–131.

Abstract—Medicinal plant classification is crucial to preserve traditional knowledge and formulate natural medicine with fewer side effects. Here, we propose an automatic medicinal plant classification using deep learning techniques. Our approach utilizes convolutional neural networks (CNNs) to classify over different types of medicinal plants based on leaf images in an efficient manner. The model classifies different medicinal plants based on leaf images. The image processing system uses the threshold method in order to remove unwanted pixels to have a clean dataset to be processed by the CNN. The model not only categorizes plant species but also offers elaborate information regarding their medicinal benefits, making it easy for users to determine the indigenous medicinal application of plants. Moreover, the model examines combinations of plants to propose possible synergies among various medicinal plants to assist in herbal medicine preparation. It offers knowledge on how combining certain plants can increase therapeutic efficacy, leading to a more effective and synergistic natural medicine. This system is anticipated to fill the knowledge gap on medicinal plants and their uses by modern generations and therefore promote the use of traditional remedies in contemporary healthcare practices. Keywords—Medicinal Plants, Plants identification, ResNet (Residual Neural Network), Feature Extraction.