DeepMetricdeep metric learning metricLearningmetric learning learningDeepdeep metric learning isMetricmetric learning a classClass of deepLearningdeep learning learningDeepdeep learning techniques focused on learning an embeddingEmbedembedding space where similar data pointsDatadata point are mapped close together while dissimilar points are pushed apart. Utilizing neural networks, these methods transform high-dimensional input data into a lower-dimensional space optimized for a chosenMetricdistance metric distance metricDistancedistance metric. This learned embeddingEmbedembedding facilitates tasksTask such as classificationClassification and retrieval by emphasizing meaningful relationships between data pointsDatadata point. This chapter provides a thorough overview ofMetricmetric learning deep Metricdeep metric learning metricLearningmetric learning learningDeepdeep metric learning approaches, beginning with foundational models such as reconstruction and Denoisingdenoising autoencoder denoising autoencodersAutoencoderdenoising autoencoder. It then explores methods that leverage classificationClassification loss functionsFunction to guideMetricmetric learning metric learningLearningmetric learning, followed by a detailed discussion on Siamese neural networks designed for pairwise similaritySimilarity learning. The chapter concludes with an explanation of self-supervised learningSelf-supervised learning (SSL) and its key techniques, highlighting recent advances inMetricmetric learning metric learningLearningmetric learning without explicit labels.