Multimedia Systems Laboratory

Multimedia Signal Processing Laboratory

Laboratory Supervisor

Hsi-Chin Hsin

Laboratory Introduction

Convolutional neural networks are one of the major research areas of this laboratory. Governments around the world continue to invest substantial manpower and resources in the hope of mastering key technologies as early as possible and creating new opportunities for technological leadership. Convolutional neural networks generally impose restrictions on the size of input images. For images of different sizes, a common approach is to resize or crop the images before input. Although this method is simple, it may damage or incorrectly crop important parts of objects. By modifying the architecture of convolutional neural networks using content-aware image resizing techniques, image recognition performance can be improved. This laboratory applies content-aware adaptive pooling techniques to extract high-quality feature points and enhance the performance of convolutional neural networks.

Interior of the Multimedia Signal Processing Laboratory

Click Num: