@article{ART003130927},
author={Se-Gwon Cheon, and Hyuk-Jin Shin and Seung-Hwan Bae},
title={3D Object Detection via Multi-Scale Feature Knowledge Distillation},
journal={Journal of The Korea Society of Computer and Information},
issn={1598-849X},
year={2024},
volume={29},
number={10},
pages={35-45},
doi={10.9708/jksci.2024.29.10.035}
TY - JOUR
AU - Se-Gwon Cheon,
AU - Hyuk-Jin Shin
AU - Seung-Hwan Bae
TI - 3D Object Detection via Multi-Scale Feature Knowledge Distillation
JO - Journal of The Korea Society of Computer and Information
PY - 2024
VL - 29
IS - 10
PB - The Korean Society Of Computer And Information
SP - 35
EP - 45
SN - 1598-849X
AB - In this paper, we propose Multi-Scale Feature Knowledge Distillation for 3D Object Detection (M3KD), which extracting knowledge from the teacher model, and transfer to the student model consider with multi-scale feature map. To achieve this, we minimize L2 loss between feature maps at each pyramid level of the student model with the correspond teacher model so student model can mimic the teacher model backbone information which improves the overall accuracy of the student model. We apply the class logits knowledge distillation used in the image classification task, by allowing student model mimic the classification logits of the teacher model, to guide the student model to improve the detection accuracy. In KITTI (Karlsruhe Institute of Technology and Toyota Technological Institute) dataset, our M3KD (Multi-Scale Feature Knowledge Distillation for 3D Object Detection) student model achieves 30% inference speed improvement compared to the teacher model.
Additionally, our method achieved an average improvement of 1.08% in 3D mean Average Precision (mAP) across all classes and difficulty levels compared to the baseline student model. Furthermore, when integrated with the latest knowledge distillation methods such as PKD and SemCKD, our approach achieved an additional 0.42% and 0.52% improvement in 3D mAP, respectively, further enhancing performance.
KW - Knowledge distillation;Multi-scale feature map;Model compression;3D object detection;Deep learning
DO - 10.9708/jksci.2024.29.10.035
ER -
Se-Gwon Cheon,, Hyuk-Jin Shin and Seung-Hwan Bae. (2024). 3D Object Detection via Multi-Scale Feature Knowledge Distillation. Journal of The Korea Society of Computer and Information, 29(10), 35-45.
Se-Gwon Cheon,, Hyuk-Jin Shin and Seung-Hwan Bae. 2024, "3D Object Detection via Multi-Scale Feature Knowledge Distillation", Journal of The Korea Society of Computer and Information, vol.29, no.10 pp.35-45. Available from: doi:10.9708/jksci.2024.29.10.035
Se-Gwon Cheon,, Hyuk-Jin Shin, Seung-Hwan Bae "3D Object Detection via Multi-Scale Feature Knowledge Distillation" Journal of The Korea Society of Computer and Information 29.10 pp.35-45 (2024) : 35.
Se-Gwon Cheon,, Hyuk-Jin Shin, Seung-Hwan Bae. 3D Object Detection via Multi-Scale Feature Knowledge Distillation. 2024; 29(10), 35-45. Available from: doi:10.9708/jksci.2024.29.10.035
Se-Gwon Cheon,, Hyuk-Jin Shin and Seung-Hwan Bae. "3D Object Detection via Multi-Scale Feature Knowledge Distillation" Journal of The Korea Society of Computer and Information 29, no.10 (2024) : 35-45.doi: 10.9708/jksci.2024.29.10.035
Se-Gwon Cheon,; Hyuk-Jin Shin; Seung-Hwan Bae. 3D Object Detection via Multi-Scale Feature Knowledge Distillation. Journal of The Korea Society of Computer and Information, 29(10), 35-45. doi: 10.9708/jksci.2024.29.10.035
Se-Gwon Cheon,; Hyuk-Jin Shin; Seung-Hwan Bae. 3D Object Detection via Multi-Scale Feature Knowledge Distillation. Journal of The Korea Society of Computer and Information. 2024; 29(10) 35-45. doi: 10.9708/jksci.2024.29.10.035
Se-Gwon Cheon,, Hyuk-Jin Shin, Seung-Hwan Bae. 3D Object Detection via Multi-Scale Feature Knowledge Distillation. 2024; 29(10), 35-45. Available from: doi:10.9708/jksci.2024.29.10.035
Se-Gwon Cheon,, Hyuk-Jin Shin and Seung-Hwan Bae. "3D Object Detection via Multi-Scale Feature Knowledge Distillation" Journal of The Korea Society of Computer and Information 29, no.10 (2024) : 35-45.doi: 10.9708/jksci.2024.29.10.035
