Few-Shot Learning (FSL) is a sub-area of machine learning that focuses on classifying new data with only a few training samples, which is particularly useful in computer vision tasks where large datasets are often unavailable. The article explores various FSL variations, including N-Shot, One-Shot, and Zero-Shot Learning, with the latter allowing for classification without any training examples. Different approaches to FSL include data-level and parameter-level strategies, with Meta-Learning being a key concept, where algorithms learn to improve performance across tasks. The article discusses several Meta-Learning algorithms for Few-Shot image classification, such as Model-Agnostic Meta-Learning (MAML), Matching Networks, Prototypical Networks, and Relation Networks, each offering unique methods for handling limited data scenarios. Few-Shot Object Detection is also addressed, highlighting the YOLOMAML algorithm, which combines YOLOv3 with MAML for effective object detection with minimal samples. Despite its challenges, FSL is a rapidly developing field with significant potential in areas where data is scarce.