Enhancing Deep Learning Model Performance through Data Generation Strategies: Classification of Olive Tree Diseases

Hicham El Akhal; Abderrahim Waga; Aissa Ben Yahya; Abdelbaki El Belrhiti El Alaoui

cs.AIOct 26, 2026

Enhancing Deep Learning Model Performance through Data Generation Strategies: Classification of Olive Tree Diseases

Authors: Hicham El Akhal, Abderrahim Waga, Aissa Ben Yahya, Abdelbaki El Belrhiti El Alaoui

Organizations: Computer Science and Applications Laboratory, Faculty of Sciences, Moulay Ismail University of Meknes, PB 11201 Zitoune, Meknes, 50000, Morocco . · Euromed University of Fez, Fez, Morocco .

Abstract

Deep learning models are highly effective in a wide range of domains, particu larly in plant disease diagnosis. However, their performance is often limited by challenges such as insufficient data and class imbalance. In our work, we propose a comprehensive data generation strategy to overcome these limitations. First, we apply various data generation techniques, including traditional augmentation methods, CutMix, MixUp, and synthetic image generation using Deep Convolu tional Generative Adversarial Networks (DCGANs), to create new data samples. Second, we exploit this generated data in three scenarios: (S1) balancing class distribution in the original dataset, (S2) increasing the number of samples per class, and (S3) expanding the dataset while maintaining balanced classes. Third, we evaluate the impact of these data distributions (S1, S2, S3) on model perfor mance using the EfficientNetV2B0 architecture with pre-trained and fine-tuned weights. The experimental results show a substantial improvement in classifica tion accuracy, emphasizing the critical role of data diversity and class balance. These findings highlight the effectiveness and adaptability of the proposed strat egy of data generation and distribution for enhancing dataset quality, offering promising applications in plant disease detection and precision agriculture.

Figures & tables

Study	Methodology	Dataset	Performance
Tabbakh and Barpanda (2023)	Hybrid model (TLMViT) with transfer learning	Wheat disease dataset	99.86% accuracy
Mhala et al. (2025)	Data augmentation, L2 regularization, transfer learning	Potato disease dataset	81.31% accuracy
Singh and Sanodiya (2023)	Zero-shot transfer learning	Tomato	-
Sathya et al. (2025)	Conditional GANs	multiple plant	98.78% accuracy
Wu et al. (2023)	ConvNeXt with attention and random masking	Soybean leaf disease dataset	85.42% accuracy
Karthikeyan and Raja (2023)	DenseNet-based CBIR (DTLDN-CBIRA)	multiple plant	100% accuracy, 81.90% recall, 89.90% F1-score

Table 1: Summary of Related Work on Data Generation and Augmentation Strategies for Classification

Study	Methods	Dataset	Accuracy	Data Generation Type
Cruz et al. (2017)	Computer vision-based deep learning	Olive leaf (1 disease + healthy)	98.60 ± 1.47%	None
Uğuz and Uysal (2021)	Deep learning (with augmentation)	Olive leaf (2 diseases + healthy)	95%	Traditional Augmentation
Alshammari et al. (2022b)	Deep learning + genetic algorithm	Olive leaf (2 diseases + healthy)	96% (multi-class), 98% (binary)	Traditional Augmentation
Alshammari et al. (2022a)	CNN + Vision Transformer	Olive leaf (2 diseases + healthy)	97% (binary), 96% (multi-class)	Traditional Augmentation
Ksibi et al. (2022)	MobiRes-Net (ResNet50 + MobileNet)	UAV-acquired images (5,400 samples)	97.08%	Traditional Augmentation
Alshammari et al. (2023)	Optimized ANN + WOA	Olive leaf (2 diseases + healthy)	98%	Traditional Augmentation

Table 2: Summary of Studies on Olive Disease Classification

Plant	Class	DataSet A	DataSet B (Test)
Olive	Healthy	1292	471
	Aculus Olearius	712	384
	Olive Peacock Spot	1716	703
	Olive Leaf Scorch	68	43
	Sooty mold	205	56
	Deficiency	206	56

Table 3: Image count per class in the dataset

Figure 2: Example images depicting various categories of diseases affecting olive leaves.

Figure 3: The architecture proposed for the classification.

Process	Approximate Duration
Data Generation using DCGAN (Generated Data)	3 months
Traditional Data Augmentation (e.g., Flip, Rotation)	1 hour
Data generation using MixUp and CutMix	1 hour
Selection of High-Quality (Generated Data) using Frechet Inception Distance (FID)	2 months
Training of Classification Models (EfficientNetV2B0)	30 minutes per experiment (see Table 5 )

Table 4: Computational Costs Associated with the Proposed Methodology

	No Aug.	Traditional Aug.		MixUp		CutMix		DCGAN		All Comb.
	Acc.	Acc.	Diff.	Acc.	Diff.	Acc.	Diff.	Acc.	Diff.	Acc.	Diff.
S1	–	91.56	+0.57	91.50	+0.51	91.90	+0.91	92.24	+1.25	93.09	+2.10
S2	–	90.76	-0.23	91.73	+0.74	90.71	-0.28	91.33	+0.34	93.48	+2.49
S3	–	90.76	-0.23	91.10	+0.11	92.29	+1.30	91.39	+0.40	90.93	-0.06
Baseline	90.99	–	–	–	–	–	–	–	–	–	–

Table 5: Test accuracy on the olive disease dataset using various data generation techniques.

	Data Generation	Precision	Recall	F1-Score	G-mean	MCC
Baseline	None	92.14	88.04	89.63	87.66	87.53
S1	Traditional Aug.	92.38	88.67	89.89	87.99	88.35
	Mixup	92.05	91.50	91.48	89.19	88.46
	Cutmix	92.17	91.90	91.91	89.25	88.77
	DCGAN	93,66	93,09	93,14	90,55	90,55
	All Combination	93.89	90.70	91.89	90.37	90.15

Table 6: Evaluation scores on the test dataset (%).

Figure 5: Ranking of Training Data Distributions Utilized in this Study for Model Training.

Synthetic Data Augmentation

Synthetic data augmentation creates or modifies training examples to improve learning when real data are scarce, imbalanced, costly, or sensitive.

See all 456 papers

A Data-Centric Review of Plant Disease Datasets: Taxonomy, Critical Analysis, Environmental Variability, and Implications for Precision Agriculture
Oct 5, 2026Aamir Hilal, Shabir Ahmad Sofi, Neeraj GoelClass-Imbalanced Learning Plant Disease Classification
Department of Information Technology, National Institute of Technology, Hazratbal Srinagar 190006, Jammu & Kashmir, India · Department of Computer Science and Engineering, Indian Institute of Technology, Ropar Rupnagar 140001, Punjab, India
An Ensemble Deep Learning Approach for Reliable and Scalable Lemon Leaf Disease Classification
Jun 12, 2026Shayan Abrar, Sudeepta Mandal, Abdul Awal Yasir +4Plant Disease Classification Explainable Image Classification
Dept. of CSE American International University-Bangladesh Dhaka, Bangladesh · Dept. of CSE East West University Dhaka, Bangladesh · Dept. of CSE North South University Dhaka, Bangladesh
Synthetic Data Generation for Long-Tail Medical Image Classification: A Case Study in Skin Lesions
May 4, 2026Jiaxiang Jiang, Mahesh Subedar, Omesh TickooSkin Lesion Classification Synthetic Data Augmentation
Intel · Google

Oct 5, 2026J/K move · Enter open · S save

A Data-Centric Review of Plant Disease Datasets: Taxonomy, Critical Analysis, Environmental Variability, and Implications for Precision Agriculture

Aamir Hilal, Shabir Ahmad Sofi, Neeraj Goel

Department of Information Technology, National Institute of Technology, Hazratbal Srinagar 190006, Jammu & Kashmir, India · Department of Computer Science and Engineering, Indian Institute of Technology, Ropar Rupnagar 140001, Punjab, India

Class-Imbalanced Learning Plant Disease Classification Precision Agriculture Cross-Dataset Generalization

Open paper AskPDF arXiv

Despite rapid advances in artificial intelligence, reliable real-world plant disease detection remains a persistent challenge. Visual and deep learning approaches have shown promising results, but their deployment under field conditions remains limited. A key bottleneck is the reliance on laboratory-generated datasets that lack environmental diversity, realistic backgrounds, and balanced class distributions, resulting in poor generalization. In contrast, datasets collected directly from agricultural environments capture natural variability and better reflect challenges faced by farmers across regions. This review presents a critical analysis of visual and deep learning approaches for plant disease detection, with emphasis on plant disease datasets. It establishes a taxonomy based on acquisition setting, accessibility, plant diversity, disease composition, class structure, and imbalance severity, and examines their implications for model generalization and real-world deployment. A comparative analysis of laboratory and real-field datasets identifies critical gaps that hinder disease detection. The review further analyzes how multi-level dataset imbalance, including intra-class, inter-crop, and cross-dataset imbalance, and limited environmental variability affect model performance and robustness, an area insufficiently examined in existing surveys. Beyond image-based approaches, it highlights the importance of integrating environmental parameters such as temperature, humidity, and leaf wetness with image data to improve prediction under dynamic field conditions. Finally, the review identifies key challenges, research gaps, and future directions concerning dataset construction, environmental variability, structural imbalance, standardization, and multimodal disease monitoring. It provides a foundation for developing next-generation multimodal frameworks for precision agriculture.

Abstract

Synthetic Data Augmentation

Explore similar work

A Data-Centric Review of Plant Disease Datasets: Taxonomy, Critical Analysis, Environmental Variability, and Implications for Precision Agriculture

An Ensemble Deep Learning Approach for Reliable and Scalable Lemon Leaf Disease Classification

Synthetic Data Generation for Long-Tail Medical Image Classification: A Case Study in Skin Lesions