| Title |
Knowledge Storage Distillation for Dense Object Detection via Non-maximum Suppression |
| Authors |
손수호(Suho Son) ; 송병철(Byung Cheol Song) |
| DOI |
https://doi.org/10.5573/ieie.2025.62.1.95 |
| Keywords |
Dense object detection; Knowledge distillation; Deep neural network compression; Non-maximum suppression |
| Abstract |
Recently, many studies have proposed knowledge distillation (KD) frameworks for object detection. However, these frameworks have not considered the inefficiencies that arise when inputting images into the teacher detector. To address this inefficiency, this paper proposes a method that pre-selects and stores only the key features from the teacher detector's feature maps using Non-Maximum Suppression (NMS), and reuses them. Despite having a similar training speed to the student model, the proposed method demonstrates superior performance compared to existing state-of-the-art knowledge distillation frameworks. |