# An Ensemble of Convolutional Neural Networks to Detect Foliar Diseases in Apple Plants

**Authors:** Kush Vora, Dishant Padalia  
**arXiv:** [2210.00298](https://arxiv.org/abs/2210.00298)  
**Date:** 2022-10-01  
**Summary page:** https://www.summarizepaper.com/en/paper/2210.00298v1/

## Key points

- Apple diseases can have a significant impact on resource loss and human/animal health.
- Early diagnosis is crucial for effective plant health management and risk reduction.
- Manual scouting and analysis of plant leaves for disease symptoms can lead to delayed diagnosis and misjudgments.
- An ensemble system consisting of three different architectures (Xception, InceptionResNet, and MobileNet) is proposed to overcome limitations.
- The system is trained on the Plant Pathology 2021 dataset, which contains images of apple plant leaves affected by various diseases.
- The developed model can detect five different types of apple plant diseases.
- Xception demonstrates the best performance among individual models with an F1 score of 0.88.
- Different combinations of models are ensembled to enhance overall performance.
- The ensemble achieves remarkable results with an accuracy of 0.8731, precision of 0.8905, recall of 0.9100, and F1-score of 0.9001.
- Comparing results with existing systems in the literature is challenging due to variations in datasets and types of diseases considered.
- Convolutional neural networks offer valuable insights into early detection and classification of apple plant diseases.
- Effective crop monitoring plays a crucial role in minimizing crop losses caused by bacterial infections in agricultural crops like apples.
- The proposed automated system provides a more efficient alternative to visual inspection alone.
- Convolutional neural networks can be implemented in real-time settings to assist farmers in managing their apple plantations effectively.

## Summary

This study addresses the issue of apple diseases and their potential impact on resource loss and human/animal health. The researchers emphasize the importance of early diagnosis to effectively manage plant health and minimize associated risks. However, the traditional approach of manually scouting and analyzing plant leaves for disease symptoms often leads to delayed diagnosis and misjudgments. To overcome these limitations, an ensemble system consisting of three different architectures (Xception, InceptionResNet, and MobileNet) is proposed by training them on a publicly available dataset called Plant Pathology 2021. This dataset contains images of apple plant leaves affected by various diseases. The developed model is capable of detecting five different types of apple plant diseases. Performance evaluation using a large testing set comprising 3,726 images shows that Xception demonstrates the best performance with an F1 score of 0.88 among the individual models. Furthermore, different combinations of these models are ensembled to enhance overall performance. The ensemble comprising Xception, InceptionResNet, and MobileNet achieves remarkable results with an accuracy of 0.8731, precision of 0.8905, recall of 0.9100, and F1-score of 0.9001. It should be noted that comparing these results with existing systems in the literature is challenging due to variations in datasets and types of diseases considered by previous studies. However, this research provides valuable insights into using convolutional neural networks for early detection and classification of apple plant diseases. The authors conclude that effective crop monitoring plays a crucial role in minimizing crop losses caused by bacterial infections in agricultural crops like apples. The proposed automated system offers a more efficient alternative to conventional methods based on visual inspection alone. Overall, this study highlights the potential benefits of utilizing convolutional neural networks for early detection and management of apple plant diseases through an ensemble system. The results show promising performance and can be implemented in real-time settings to assist farmers in effectively managing their apple plantations.