<![CDATA[This research aimed to develop a Convolutional Neural Network (CNN) model for automatic object color classification using MobileNetV2. To determine the optimal configuration, the training process adjusted several hyperparameters, with particular focus on identifying the most suitable learning rate. The dataset consisted of 3,212 images grouped into five color categories: red, green, blue, random (including yellow, orange, and brown), and none (no object detected). Data augmentation techniques were applied to enhance the variety and robustness of the dataset. The model was trained using the Adam optimizer alongside the categorical crossentropy loss function, with various learning rate settings tested during training. Evaluation results showed that the model worked best with a learning rate of 0.0001 and a batch size of 32, with an average accuracy of 94%. To display prediction results in real time, the top-performing model was integrated into a graphical user interface (GUI). These findings demonstrate the effectiveness of the MobileNetV2-based CNN model in recognizing object colors and highlight its suitability for integration into real-time industrial sorting applications]]>
Copyrights © 2025
