Classifying Species of Plant through Deep Learning

Abstract

Planet leaf identification for medicinal usage is a slow and human error-prone task based on expert manual identification. Classifying plant leaves according to their type and medicinal use is challenging based on current research that assigns a new approach through DL& (AI). The suggested system employs a (CNN) in filtering the visual characteristics of the planet leaves and (NLP) techniques in extracting information from the history of the leaf and medicine applied in disease treatment. The suggested system achieved an accuracy level of 95% in classifying the type of leaves and medical use. The outcome justifies the suggested system\'s effectiveness, accuracy, and potential in redefining plant leaf identification and classification for medicinal usage. The current paper expounds on the suggested system, methodology, result analysis, and conclusion.

Introduction

1. Introduction

Plant leaves have long been used in traditional medicine, but manual classification is slow, error-prone, and highly dependent on expert knowledge. A major challenge is the lack of standardized, automated methods for identifying leaves and their medicinal uses. This paper proposes an AI-driven deep learning system to automate the classification of plant leaves and predict their therapeutic properties using both visual and textual data.

2. Objectives & Problem Statement

• Existing manual and chemical-based methods for identifying medicinal plants are:

  • Time-consuming

  • Expensive

  • Require specialized skills

• Most current AI systems rely only on visual data, neglecting important textual information (e.g., historical medicinal uses).

• There is a need for a precise, scalable, and generalizable method for classifying plant leaves and understanding their medical applications.

3. Proposed Solution

The system combines:

• Convolutional Neural Networks (CNNs) for extracting visual features from leaf images.

• Natural Language Processing (NLP) techniques for analyzing associated textual data (e.g., origin, medicinal uses, treated diseases).

• A feature fusion technique (e.g., concatenation) integrates visual and textual features.

• Machine learning classifiers like Support Vector Machines (SVM) and Random Forests (RF) predict the leaf type and its medicinal applications.

4. Methodology

The process is divided into four key stages:

1. Dataset Collection & Preprocessing

   • Images and associated medicinal text data are collected from online repositories and journals.

   • Preprocessing involves image resizing and normalization; text cleaning includes tokenization, stemming, and stopword removal.

2. Visual Feature Extraction (CNN)

   • A CNN model extracts distinguishing visual traits like shape and texture from the plant leaf images.

3. Textual Feature Extraction (NLP)

   • Texts are processed with NLP techniques like Named Entity Recognition and dependency parsing.

   • Extracted text features are converted into vectors for integration.

4. Feature Fusion & Classification

   • Visual and textual features are fused and passed to classifiers (SVM, RF) to predict both the plant type and its medicinal use.

   • Evaluation metrics include accuracy, precision, recall, and F1 score.

5. Related & Existing Work

• Past research focused mainly on visual features using CNNs (e.g., Wang 2020, Chung 2022).

• Some hybrid models (e.g., Li et al. 2021) incorporated morphological analysis but ignored historical medicinal context.

• Current systems often:

  • Lack textual context

  • Are not generalizable to novel or extraterrestrial plant species

  • Use earth-bound data only and struggle in simulated or alien environments

6. Results

• The proposed system outperforms existing models in classification accuracy and completeness of information.

• Integration of image and text improves:

  • Consistency

  • Diagnostic accuracy

  • Insight into medicinal use

• Feature fusion provides a richer representation of each plant leaf.

• The model is validated using a labeled dataset and shows strong performance in recognizing leaf type and therapeutic value.

Conclusion

In this proposal, we introduce a completely new idea for the classification of plant leaves and some of their medicinal uses based on the visual as well as textual features of a plant. The proposed solution is the integration of artificial intelligence and deep learning techniques, i.e., the application of a computer vision model with a CNN for feature extraction, the application of a natural language processing (NLP) technique for textual feature extraction, and a feature fusion technique for visual and textual data. The proposed solution is precise, efficient, and comprehensive in the identification of plant leaves and their medicinal uses. Experiments done using a dataset of planet leaf pictures and their respective textual descriptions will be presented and examined. The outcomes will be described in the context of the newly proposed solution and the already existing one, stressing the positives and negatives of each method. To put it on another level, we consider that such an implementation will not only give the field of plant-based medication an unprecedented dimension. Still, we will also become the foundation of the health and quality of life of people by it. The suggested solution is the one correctly identifying the planet leaves and their medicinal use. In this manner, they can start a process that will standardize the classification of plant-based medicines, which will be the use of plant-based medicines. The previous work can include the extension of the proposed solution to other areas of plant-based medicines, e.g., plant disease identification and classification, and the extension of the proposed solution to smartphones and other platforms for the sake of facilitating mass application by the population.

References

[1] https://www.researchgate.net/publication/329477 251_A_Leaf_Recognition_Approach_to_Plant_C lassification_Using_Machine_Learning [2] https://ieeexplore.ieee.org/document/8675114 [3] https://www.sciencedirect.com/science/article/abs/pii/S1537511015001373 [4] https://www.frontiersin.org/journals/plantscience/articles/10.3389/fpls.2023.1286088/full [5] https://pmc.ncbi.nlm.nih.gov/articles/PMC10375 553/ [6] https://www.propulsiontechjournal.com/index.ph p/journal/article/download/5988/3959/10290 [7] https://www.frontiersin.org/journals/plantscience/articles/10.3389/fpls.2023.1286088/full [8] Lecun, Y., Bottou, L., Bengio, Y., & Haffner, P. (1998). Gradient-based learning applied to document recognition. Proceedings of the IEEE, 86(11), 2278- 2324.https://ieeexplore.ieee.org/document/6795954 [9] INetCache/IE/KM1X2QQZ/ajol-filejournals_333_articles_217008_submission_proof_217008-3973-534245-1-10-20211106[1].pdf https://doi.org/10.33851/JMIS.2019.6.2.49

Copyright

Copyright © 2025 K. Satvika, D. Gayatri, B. Tejaswini, Sivaiah Borra , Dr. G. Sudhakar , Dr. J. Rajeshwar. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.