Shot Class

Shot class, specifically few-shot class incremental learning (FSCIL), focuses on enabling machine learning models to efficiently learn new categories from limited data without forgetting previously acquired knowledge. Current research emphasizes adapting pre-trained vision models (like Vision Transformers and CLIP) through techniques such as fine-tuning specific layers, incorporating contrastive learning, and utilizing graph neural networks to leverage relationships between samples and classes. This area is crucial for developing more robust and adaptable AI systems capable of continuous learning in real-world scenarios with limited labeled data, impacting fields like robotics and personalized AI.