Automatic Flower Classification Using CNN Models

Abstract

This paper illustrates the flower categorization methods utilizing deep structured learning for the recognition of different flower types through images. Pre-trained MobileNetV2 is adopted to classify various flower types like daisy, dandelion, rose, sunflower, and tulip with precision and efficiency. Image analysis methods like photo resizing, normalization, and augmentation have been considered to increase the results of the system. The skilled Neural Networks model is coupled with an easy-to-use interface where users can feed flower images to get the predicted result along with its confidence score. The proposed system delivers robust performance at low cost, making it feasible for real world applications.

Introduction

The text presents a deep learning-based flower classification system developed using the MobileNetV2 convolutional neural network architecture to automatically identify flower species from images. The project aims to simplify botanical identification by applying Artificial Intelligence and computer vision techniques to classify flowers such as daisy, dandelion, rose, sunflower, and tulip. A web-based interface allows users to upload flower images and receive instant classification results with prediction confidence.

The literature review discusses the evolution of image classification models, including AlexNet, VGG16, ResNet, MobileNetV2, and SqueezeNet. Earlier CNN architectures achieved high accuracy but required significant computational resources. Lightweight models such as MobileNetV2 and SqueezeNet were developed to reduce memory usage and processing requirements while maintaining competitive performance. Among these, MobileNetV2 is selected because of its balance between accuracy, speed, and efficiency, making it suitable for real-time flower recognition applications.

The system uses several deep learning techniques, including Convolutional Neural Networks (CNNs) for automatic feature extraction, transfer learning to reuse previously learned image features, data augmentation to improve model generalization, and image preprocessing techniques such as resizing and normalization to enhance input quality. MobileNetV2 analyzes flower characteristics including color, texture, petal shape, and structural patterns to classify images into different categories.

The proposed system architecture consists of multiple stages: image upload by the user, preprocessing of the input image, feature extraction using MobileNetV2, classification of flower categories, and display of prediction results through a user-friendly web interface. The lightweight design enables fast predictions with reduced computational complexity, making the system suitable for applications such as education, agriculture, botanical research, and intelligent image recognition.

The model performance was evaluated using accuracy, loss, precision, recall, and F1-score metrics. Experimental results demonstrate effective classification performance with:

• Accuracy: 92%
• Loss: 0.25
• Precision: 91%
• Recall: 90%
• F1-Score: 90.5%

The results indicate that MobileNetV2 successfully identifies flower species with high efficiency while requiring fewer computational resources compared to traditional deep learning models. Image preprocessing and transfer learning contributed significantly to improving classification accuracy and reducing training complexity.

The conclusion highlights that the developed flower classification system demonstrates the effectiveness of deep learning in automated plant identification. Future improvements include expanding the number of flower categories, using larger datasets, experimenting with advanced architectures such as EfficientNet, ResNet, and Vision Transformers, integrating the system into mobile applications, and adding additional features such as plant disease detection, multilingual support, voice assistance, and cloud-based prediction. These enhancements can increase the usefulness of the system in agriculture, education, environmental monitoring, and biodiversity conservation.

Conclusion

The flower classification system developed in this project demonstrates the effective usage of neural network techniques for automated image recognition. By using the MobileNetV2 architecture, the system successfully classifies different flower categories with good accuracy and reduced computational complexity. The integration of image preprocessing, feature extraction, and classification techniques helped improve the overall performance of the model. The web-based interface further enhances user interaction by allowing users to upload images & instantly receive predicting outputs by means of confidence level and performance graphs. The model main features how ML(machine learning) and computer vision can simplify botanical identification tasks that normally require manual observation and expertise. The obtained output shows that neural networks system can efficiently recognize flower patterns, colors, and textures even under different image conditions. In addition, the lightweight nature of MobileNetV2 makes the model suits for real-time applications and devices with limited processing power. In future developments, the project can be expanded by increasing the number of flower categories and using larger datasets to improve classification accuracy. ADLM(Advanced deep learning models) such as EfficientNet, ResNet, or Vision Transformers can also be executed for excellence characteristic filter and performance. The system may further be integrated with mobile applications so users can identify flowers directly using smartphone cameras. Additional features such as disease detection in plants, multilingual support, voice assistance, and cloud-based prediction systems can also be included. These future improvements can make the system more useful in agriculture, education, environmental monitoring, and biodiversity conservation.

References

[1] Y. LeCune, Y. Bengeio, and G. Hineton, “Deep Learning Model,” Nature, vol.. 521, no. 7573, p.p. 436–444, 2015. [2] M. Sandler, A. Howard, M. Zhu, A. Zhmoginov, and L. Chen, “MobileNetV2: Inverted Residuals and Linear Bottlenecks,” in case of the IEEE Confernce on System Vision and Pattern Identification (CVPR), 2018. [3] A. Krizhevsky, I. Sutkever, and G. Hineton, “ImageNet Clasification with Deep Convolutional Neural Networks(DNN)” in progress in Neueral Information, Processing Model, 2013. [4] K. Simaonyan and A. Ziserman, “Very Deep Convolutional(VDC) Networks for Greater-Scale image Identification,” Internatonal Conferences on Learning Representations (ICLR), 2015. [5] K. Hes, X. Zhaneg, S. Reen, and J. Sune, “Neural Residual Learning for Image Recognition,” in Procedings of the IEEE Confernce on System Vision & Pattern Identification (CVPR), 2016. [6] F. N. Iandola, S. Haan, M. Moskewiczz, K. Ashreaf, W. Daly, and K. Keuytzer, “SquezeNet: AlexNet-Step Accuracy with 40x Lesser Parameters & <0.5 MB Model-Size,” arrXiv preprint arXiv:1602.07360, 2017. [7] I. Goodfellow, Y. Bengieo, and A. Courvile, Deep Learning, MITG Press, 2016. [8] R. Szeliski, Computer Vision: Algorithms and Applications, Springer, 2011. [9] A. Gurnani, V. Mavani, V. Gajjar, and Y. Khandhediya, “Flower Identification using Deep CNN,” arXiv preprint arXiv:1708.03763, 2017. [10] A. Howard et al., “Searching for MobileNetV3,” Procedings of the IEEE International Conference on Computer Vision (ICCV), 2019. [11] S. Kosankar and V. Khan, “Flower Classification using MobileNet: An Optimized Deep Learning Model,” Inter-national Publication of Advanced Research in Computer and Communication Engineering (IJARCCE), vol. 81, no. 4, 2019. [12] V. Ayumi, “Flower Object Recognition Depending on Deep neural Learning Using ResNet50 and MobileNetV2,” Journal Scientific and Applied Informatics, vol. 6, no. 3, 2022. [13] S. Navyuakala, N. Paddmaja, and P. Nelima, “A Studied on Flower-Classification-Using-Deep Learning-Techniques,” International Journal of Computer Engineering in Research Trends, vol. 10, no. 4, 2023. [14] V. Praskatama, G. F. Shidik, and A. P. Ningrum, “Comparative Study: Flower Classification using Deep Learning, SMOTE and Fine-Tuning,” Journal of Applied Informatics and Computing, vol. 8, no. 2, 2024. [15] E. Yulisara, N. Husna, D. Martin, and C. Ariesta, “Comparative Study of Convolutional Neural networks Strucures and Optimizers for Flower Image Classification,” Public Research Journal of Engineering, Data Technology and Computer Science, 2025. [16] A. I. Ramadhona, W. S. Astuti, D. Hartama, A. P. Windarto, and A. Wanto, “Enhanced Flower Image Identification by Using MobileNetV2 with Optimized Performance,” JITK Journal of Computer Science and Technology(CST), vol. 11, no. 1, 2025. [17] TensorFlow Documentation, “TensorFlow: Open Source Machine Learning Framework,” Available: https://www.tensorflow.org [18] Keras Documentation, “Keras Deep Learning API,” Available: https://keras.io [19] Flask Documentation, “Flask Web Framework for Python,” Available: https://flask.palletsprojects.com [20] NumPy Documentation, “NumPy Numerical Computing Library,” Available: https://numpy.org

Copyright

Copyright © 2026 Dr. Bhavana R Maale, Varalakshmi N. 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.