AI RESEARCH
FiGKD: Fine-Grained Knowledge Distillation via High-Frequency Detail Transfer
arXiv CS.CV
•
ArXi:2505.11897v2 Announce Type: replace Knowledge distillation (KD) is a widely adopted technique for transferring knowledge from a high-capacity teacher model to a smaller student model by aligning their output distributions. However, existing methods often underperform in fine-grained visual recognition tasks, where distinguishing subtle differences between visually similar classes is essential.