<![CDATA[This study presents the implementation of the FaceNet deep convolutional neural network model as a facial search function for a reverse image search system. The research addresses the growing need for fast and accurate face recognition in digital applications, focusing on embedding-based similarity search rather than traditional classification methods. The system employs MTCNN for face detection and alignment, followed by the FaceNet model to generate 128-dimensional facial embeddings whose distances represent the similarity between identities. The methodology includes data preprocessing, embedding extraction, distance-based matching, and system evaluation using images with identical and non-identical identities. Experimental results show that the average embedding distance for identical faces is 0.47, while non-identical faces exhibit an average distance of 0.62. The proposed system achieves an accuracy of 94% on a test set of 100 images, demonstrating its effectiveness in distinguishing facial similarities. These findings confirm that embedding-based representation using FaceNet provides a reliable foundation for facial retrieval tasks in reverse image search applications, offering high discriminative capability and operational efficiency.]]>
Copyrights © 2025
