Performance Enhancement of Deep Learning Models Through Optimization Techniques for Automated Apple Grading

Sandeep Kaur; Manoj Kumar Sachan; Ashwani Kumar Aggarwal

doi:10.29284/2dgx0h65

Authors

Sandeep Kaur Department of Computer Science and Engineering, Sant Longowal Institute of Engineering and Technology, Longowal-148106, India, Email: sandeep.sandeep.sran@gmail.com
Manoj Kumar Sachan Department of Computer Science and Engineering, Sant Longowal Institute of Engineering and Technology, Longowal-148106, India, Email: manojsachan@sliet.ac.in
Ashwani Kumar Aggarwal Department of Electrical and Instrumentation Engineering, Sant Longowal Institute of Engineering and Technology, Longowal-148106, India, Email: ashwani.ist@sliet.ac.in

DOI:

https://doi.org/10.29284/2dgx0h65

Keywords:

Automated Apple Grading; Deep Learning; Hyperparameter Optimization; Bayesian Optimization; Differential Evolution; EfficientNet; CNN–Autoencoder Fusion; Transfer Learning; Agricultural Image Classification

Abstract

Automated grading of agricultural produce requires accurate and robust visual classification, yet the performance of deep learning models is highly sensitive to hyperparameter selection and training dynamics. This study presents an optimization-driven deep learning framework for automated apple grading aimed at enhancing classification accuracy, convergence stability, and feature representation through systematic hyperparameter tuning. A dataset of 1,264 images of Royal Delicious apples categorized into four grades was used to evaluate CNN, ResNet50, EfficientNet-B0, and a CNN–Autoencoder fusion model under baseline and optimized settings. Baseline results with default hyperparameters achieved accuracies of 90.8% (CNN), 93.7% (ResNet50), and 95.1% (EfficientNet-B0), indicating moderate generalization and higher validation loss. After applying Grid Search, Bayesian Optimization, PSO, and Differential Evolution, significant performance improvements were observed. The proposed CNN–Autoencoder fusion model optimized using Differential Evolution achieved the highest accuracy of 97.3%, with precision 0.97, recall 0.96, F1-score 0.96, lowest validation loss (0.15), and fastest convergence (epoch 17). Overall, optimization improved accuracy, reduced loss, enhanced training stability, and produced more discriminative latent features, confirming the effectiveness of systematic hyperparameter optimization in automated grading systems.

References

[1] Ahmed, S. F., Alam, M. S. B., Hassan, M., Rozbu, M. R., Ishtiak, T., Rafa, N., ... &Gandomi, A. H. (2023). Deep learning modelling techniques: current progress, applications, advantages, and challenges. Artificial Intelligence Review, 56(11), 13521-13617.

[2] Saputra, N. A., Riza, L. S., Setiawan, A., & Hamidah, I. (2024). A systematic review for classification and selection of deep learning methods. Decision Analytics Journal, 12, 100489.

[3] Raihen, M. N., & Akter, S. (2024). Prediction modeling using deep learning for the classification of grape-type dried fruits. International Journal of Mathematics and Computer in Engineering, 2(1), 1-12.

[4] Jia, W., Sun, M., Lian, J., & Hou, S. (2022). Feature dimensionality reduction: a review. Complex & Intelligent Systems, 8(3), 2663-2693.

[5] Akay, B., Karaboga, D., & Akay, R. (2022). A comprehensive survey on optimizing deep learning models by metaheuristics. Artificial Intelligence Review, 55(2), 829-894.

[6] Hanifi, S., Cammarono, A., & Zare-Behtash, H. (2024). Advanced hyperparameter optimization of deep learning models for wind power prediction. Renewable Energy, 221, 119700.

[7] Bergstra, J., & Bengio, Y. (2012). Random search for hyper-parameter optimization. Journal of machine learning research, 13(2).

[8] Khan, B., Das, S., Fahim, N. S., Banerjee, S., Khan, S., Al-Sadoon, M. K., ... & Islam, A. R. M. T. (2024). Bayesian optimized multimodal deep hybrid learning approach for tomato leaf disease classification. Scientific Reports, 14(1), 21525.

[9] Jabbar, A. N., & Koyuncu, H. (2023). Deep learning and grey wolf optimization technique for plant disease detection: a novel methodology for improved agricultural health. Traitement du Signal, 40(5), 1961-1972.

[10] Abdullah, A., Amran, G. A., Tahmid, S. A., Alabrah, A., AL-Bakhrani, A. A., & Ali, A. (2024). A deep-learning-based model for the detection of diseased tomato leaves. Agronomy, 14(7), 1593.

[11] Ivanovski, T., Gulić, M., &Matetić, M. (2024). Optimization of CART Models Using Metaheuristics for Predicting Peach Firmness. Applied Sciences, 14(18), 8539.

[12] Han, B., Zhang, J., Almodfer, R., Wang, Y., Sun, W., Bai, T., ... & Hou, W. (2025). Research on innovative apple grading technology driven by intelligent vision and machine learning. Foods, 14(2), 258.

[13] Dewi, P. K., Niswar, M., & Zainuddin, Z. (2024, November). Melon Fruit Quality Classification with Hyperparameter Optimization of Random Forest Algorithm Using Grey Wolf Optimizer. In 2024 7th International Conference on Information and Communications Technology (ICOIACT) (pp. 276-281). IEEE.

[14] Ukwuoma CC, Zhiguang Q, Bin Heyat MB, Ali L, Almaspoor Z, Monday HN. Recent advancements in fruit detection and classification using deep learning techniques. Mathematical Problems in Engineering. 2022;2022(1):9210947.

[15] Naveenkumar, M., & Nandagopal, S. (2025). Adaptive hybrid segmentation combined with meta heuristic optimization in transfer learning for plant leaf disease classification. Scientific Reports, 15(1), 9838.

[16] Kaur, S. — Sachan, M. K. — Aggarwal, A. K.: Classification of royal delicious apples using hybrid feature selection and feature weighting method based on SVM classifier. Scalable Computing: Practice and Experience, Vol. 26, No. 2, 2025, pp. 940– 949.

[17] Taner, A., Mengstu, M. T., Selvi, K. Ç., Duran, H., Gür, İ., & Ungureanu, N. (2024). Apple varieties classification using deep features and machine learning. Agriculture, 14(2), 252.

[18] Khan, M. A., Lali, M. I. U., Sharif, M., Javed, K., Aurangzeb, K., Haider, S. I., ... & Akram, T. (2019). An optimized method for segmentation and classification of apple diseases based on strong correlation and genetic algorithm based feature selection. IEEE Access, 7, 46261-46277.

[19] Kour, V. P., & Arora, S. (2019). Particle swarm optimization based support vector machine (P-SVM) for the segmentation and classification of plants. IEEE Access, 7, 29374-29385.

[20] Agarwal, M., Kaliyar, R. K., & Gupta, S. K. (2022, July). Differential Evolution based compression of CNN for Apple fruit disease classification. In 2022 International Conference on Inventive Computation Technologies (ICICT) (pp. 76-82). IEEE.

[21] Kunduracioglu, I. (2024). Cnn models approaches for robust classification of apple diseases. Computer and Decision Making: An International Journal, 1, 235-251.

[22] Zhang, X., Li, H., Sun, S., Zhang, W., Shi, F., Zhang, R., & Liu, Q. (2023). Classification and identification of apple leaf diseases and insect pests based on improved ResNet-50 model. Horticulturae, 9(9), 1046.

[23] Nashrulloh, M. H., &Subarkah, A. (2025). Fruit Image Classification Using CNN With EfficientNet Architecture for Visual Education. Journal of Intelligent Systems Technology and Informatics, 1(2), 67-71.

[24] Tian, X., Huang, L., Zhai, M., Zhang, M., Hu, P., Li, M., & Ren, L. (2025). Non-destructive prediction of apple SSC/TAC and firmness based on multilayer autoencoder and multilayer perceptron. Intelligence & Robotics, 5(1), 181-201.