Automatic image recognition is a convenient option for labeling and categorizing fruits and vegetables in supermarkets. This paper proposes the design and implementation of an automatic classification system for Colombian fruits, by training a convolutional neural network. A database was created to train and test the system, which consisted of 4980 images, labeled in 22 classes, each corresponding to pictures of the same kind of fruit, trying to reproduce the variability of a real case scenario with occlusions, different positions, rotations, lightings, colors, etc., and the use of bags. On-training data augmentation was used to further increase the robustness of the model. Additionally, transfer learning was implemented by taking the parameters of a pretrained model used for fruit classification as the new initial parameters of the proposed convolutional network, achieving an increase of the classification accuracy compared with the same model when trained with random initial weights. The final classification accuracy of the network was 98.12% which matches the scores achieved on previous works that performed fruit classification on less challenging datasets. Considering top-3 classification we report an accuracy of 99.95%.
A methodology for the automatic recognition of Colombian car license plates using convolutional neural networks is proposed. One of the biggest challenges when using onvolutional neural network is the demand for large amounts of samples for training. In this work, we show that if we do not have enough images of vehicle license plates to carry out the training, we can complement it with databases of letters and numbers that are not extracted from cars. The network was trained with the Chars74k database and images of characters extracted from plates of Colombian automobiles. The Chars74k contains approximately 74000 images of all the letters of the Spanish alphabet and all digits from 0 to 9. From chars74k database we have chosen 33849, because the Colombian plates have only uppercase letters and digits. Only 3549 (about 10% of the total) images of characters extracted manually from plates of Colombian automobiles were added. At the input of the convolutional neural network, 70% of the images were used for training, 20% for validation and 10% for testing and the resulting validation accuracy was above 99%. By making a preliminary test on Colombian plates never before used in training, a percentage of correctly recognized plates above 98% was achieved.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.