ProtoConNet: Prototypical augmentation and alignment for open-set few-shot image classification

Key Points

• Improved accuracy in open-set few-shot image classification is achieved through alignment algorithms.
• The proposed ProtoConNet uses 15% higher accuracy than traditional few-shot learning methods on standard datasets.
• Analysis focuses on prototypical networks and their augmentation to address classification challenges effectively.
• These findings support the need for innovative algorithms in a rapidly growing field of image recognition.