SPEAKERS
![]() |
Assoc. Prof. Arief Setyanto, S.Si., M.T., Ph.D. Faculty of Computer Science, Universitas Amikom Yogyakarta |
Biography
Assoc. Prof. Arief Setyanto, S.Si., M.T., Ph.D. was born in 1975 in Banyumas, Central Java, Indonesia. Currently he works as lecturer and researcher in the Universitas Amikom Yogyakarta Indonesia. He received his PhD in The School of Computer Science and Electronics Engineering (CSEE), University of Essex, Colchester The United Kingdom in 2016 under Indonesian Government scholarship. He received bachelor and Master degree from Gadjah Mada University, Yogyakarta, Indonesia in 1998, 2003 respectively. His research interest includes, video/image segmentation and understanding, object tracking, visual content metadata, information retrieval, object detection, near-edge computing, and deep learning. He can be contacted at email: arief_s@amikom.ac.id. |
ABSTRACT
“Deep Learning Model Compression for Edge AI”
Deep Learning has achieved incredible achievement in computer vision related task. However, most of deep learning model suffers from huge size. Therefore, require huge memory and computing capacity. Edge device on the other hand has potential implementation on the real world due to its tiny size and lower power requirements. Artificial intelligence on the edge implementation such as in robotics, self-driving cars, smart agriculture and many other areas are promising. Big model size hindering most deep learning model implementation on edge device. Reducing the model size become a significant task to enable Edge AI. There are many ways to reduce the size such as reducing the bit size of model weight parameters, reducing the number of parameters or transferring the knowledge to a smaller model. Balancing between the size reduction and the ability of the model to carry out computer vision task. We explore several compression techniques, their size reduction benefit and the accuracy loss. We also explore its latency benefit on real edge device, power, memory, CPU and GPU consumption.