Semi-Supervised Image Classification Using Support Vector Machines

This approach implements semi-supervised image classification using Support Vector Machines (SVM). The method from literature [1] performs supervised image feature learning from all available data (both labeled and unlabeled samples). It leverages rich categorical attributes of images to automatically generate prototype set collections from existing data. Feature learning is then applied to each prototype set, and the projected features are concatenated to form comprehensive image representations for classification. The performance is compared with traditional semi-supervised methods that determine classification boundaries based solely on unlabeled images.