| Title |
Dental Image Analysis Network Pruning via Differentiable Gate and Differential Evolution Algorithm |
| DOI |
https://doi.org/10.5573/ieie.2024.61.1.67 |
| Keywords |
Medical image analysis; Filter pruning; Deep neural network compression |
| Abstract |
Deep learning based dental image analysis algorithms show high performance, but have very high computational costs. To solve this problem, this paper proposes a method of pruning filters in a deep learning network through sigmoid-based differential gate and operators of a differential evolution algorithm. The proposed method facilitates compression rate control, and explores more sub-networks to find the optimial sub-network while minimizing performance degradation even at high compression rates. As a result, the proposed method made it possible to compress the computational cost(MACs) of the cephalometric landmark detection network up to 2.5% while minimizing performance degradation. |