Global Journal of Computer Science and Technology, G: Interdisciplinary, Volume 23 Issue 1
Table 5: Results of the final step Original & Blur Original & Contrast Training dataset 50 056 51 315 Validation dataset 12 514 12 828 Test dataset 500 500 Total 63 070 64 643 Training Accuracy (%) 98.54 97.44 Testing Accuracy (%) 97.80 93.98 Fig. 8: Training result of the ‘Contrast & Flip & Affine Transformation’ dataset Fig. 9: Training result of the ‘Flip & Zoom & Affine Transformation’ dataset V. C onclusion and F uture W orks This paper presented three contributions. The first allowed us to know the impact of data augmentation techniques namely blur, contrast, flip, noise and zoom in mango leaf diseases classification. The second is to know the best combinations between these techniques which give the best performance to the deep learning model. The last one reveals that applaying ‘affine transformation’ technique to the combination ‘Contrast & Flip’ gives the best performance to the Resnet50 CNN with an accuracy of 97.80%. This solution can be used to improve the performance of DL models for image classification with small datasets. Our future work, is to propose a dataset of mango leaf diseases with images captured in mango orchards of a sahelian country like Senegal. Applying A Combination of Data Augmentation Techniques for Mango Leaf Diseases Classification © 2023 Global Journals Global Journal of Computer Science and Technology Volume XXIII Issue I Version I 8 ( ) Year 2023 G
RkJQdWJsaXNoZXIy NTg4NDg=