Classification of Fashion-MNIST Dataset by Using Lenet-5
DOI:
https://doi.org/10.61173/8nf5y380Keywords:
LeNet-5, Fashion-MNIST, Image Recognition, Convolutional Neural NetworksAbstract
This study explores the application of LeNet-5 convolutional neural network (CNN) on the Fashion-MNIST dataset to analyze its effectiveness in complex tasks using image recognition. The Fashion-MNIST dataset, comprising 70,000 28x28 pixel clothing images across 10 distinct categories, is constituted to offer a challenging benchmark for image recognition research. In the experiment, the data is first normalized and the labels are preprocessed using the one-hot encoding technique to adjust to the model’s input specifications. The LeNet-5 network combines several fully connected, pooling, and convolutional layers before using the Softmax layer to produce the probability distribution for each category. Through a series of experiments, it is found that LeNet-5 achieves 95% training accuracy and 90% validation accuracy on the Fashion-MNIST dataset, which not only confirms its effectiveness in modern image recognition tasks, but also demonstrates its good generalization ability. The study’s findings provide new insights for the future application of traditional CNN in the field of deep learning and provide new perspectives for future applications in more complex image recognition tasks. In addition, this study also provides the possibility for the application of LeNet-5 in scenarios with limited resources or high real-time requirements, providing valuable experience and reference for subsequent research.References
[1] El-Sawy A, El-Bakry H, Loey M. CNN for handwritten arabic digits recognition based on LeNet-5. Proceedings of the International Conference on Advanced Intelligent Systems and Informatics 2016 2. Springer International Publishing, 2017: 566-575.
[2] Wang Y, Li F, Sun H, et al. Improvement of MNIST image recognition based on CNN. IOP Conference Series: Earth and Environmental Science. IOP Publishing, 2020, 428(1): 012097.
[3] Kumar A, Singh I, Singh B. Classification of MNIST Dataset Fig. 3 Epoch-wise Training and Validation Using Different CNN Architecture. Intelligent Circuits and Accuracy Progression Systems for SDG 3–Good Health and well-being, 2024: 273- The model evaluation results in a composite score on the 280. testing set. The model’s performance on the testing set [4] Kadam S S, Adamuthe A C, Patil A B. CNN model for image further demonstrates its supremacy. The loss of testing classification on MNIST and fashion-MNIST dataset. Journal of and the accuracy of testing are 0.2 and 95%, respectively, scientific research, 2020, 64(2): 374-384. and these measurements reach the field’s existing state-of- [5] Nocentini O, Kim J, Bashir M Z, et al. Image classification the-art. using multiple convolutional neural networks on the fashion- MNIST dataset. Sensors, 2022, 22(23): 9544.
[6] Bin Tareaf R, Korga A M, Wefers S, et al. Applying XAI in
[4] Summary Fashion-MNIST: Evaluating Hybrid Deep Neural Networks and This paper comprehensively evaluates the performance Capsule Networks. Proceedings of the 2024 16th International of the classical LeNet-5 CNN in complex image recogni- Conference on Machine Learning and Computing. 2024: 100- tion tasks through its application to the Fashion-MNIST 107. dataset. The results show that LeNet-5 exhibits excellent [7] Haji L M, Mustafa O M, Abdullah S A, et al. Enhanced recognition ability on the Fashion-MNIST dataset, and Convolutional Neural Network for Fashion Classification. the loss function’s continuous decrease during training Engineering, Technology & Applied Science Research, 2024, and validation, as well as the significant improvement of 14(5): 16534-16538. the training accuracy and validation accuracy, validate the [8] LEITHARDT V. Classifying garments from fashion-MNIST model’s excellent learning ability and generalization ca- dataset through CNNs. Advances in Science, Technology and pability. In addition, the test set further confirms the mod- Engineering Systems Journal, 2021, 6(1): 989-994. el’s effectiveness and reliability in practical applications [9] LeCun Y, Bottou L, Bengio Y, et al. Gradient-based learning through its excellent performance. applied to document recognition[J]. Proceedings of the IEEE, This study not only provides an effective solution for 1998, 86(11): 2278-2324. image recognition on the Fashion-MNIST dataset, but [10] Zhang J, Yu X, Lei X, et al. A novel deep LeNet-5 also provides a strong demonstration of the potential of convolutional neural network model for image recognition. LeNet-5 to be applied in modern image recognition tasks. Computer Science and Information Systems, 2022, 19(3): 1463- Although LeNet-5 is a relatively old model, its perfor- 1480. mance in this study shows that it can still be comparable [11] Wei G, Li G, Zhao J, et al. Development of a LeNet-5 gas to modern models in certain tasks. This finding opens up identification CNN structure for electronic noses. Sensors, 2019, the possibility of future applications of LeNet-5 in scenar- 19(1): 217. ios with limited resources or high real-time requirements. [12] Kayed M, Anter A, Mohamed H. Classification of garments In conclusion, this study’s findings expand the field in from fashion MNIST dataset using CNN LeNet-5 architecture. which LeNet-5 can be applied while also offering invalu- 2020 international conference on innovative trends in able knowledge and resources for future investigations communication and computer engineering (ITCE). IEEE, 2020: into increasingly intricate picture identification problems. 238-243. Future work can further explore the combination of LeN-
Downloads
Published
Issue
Section
License
Copyright (c) 2024 by the authors.
This work is licensed under a Creative Commons Attribution 4.0 International License.
