Machine learning: model compression techniques and deployment on Android platform

As Artificial Intelligence (AI) industry grows rapidly in recent years, many applications of deep learning are applied in mobile devices where resources are limited. Therefore, model compression and acceleration techniques are of great importance for achieving real-time requirements. In this study,...

ver descrição completa

Detalhes bibliográficos
Autor principal: Jin, Chengkai
Outros Autores: Jun Zhao
Formato: Final Year Project (FYP)
Idioma:English
Publicado em: Nanyang Technological University 2022
Assuntos:
Acesso em linha:https://hdl.handle.net/10356/156646
Descrição
Resumo:As Artificial Intelligence (AI) industry grows rapidly in recent years, many applications of deep learning are applied in mobile devices where resources are limited. Therefore, model compression and acceleration techniques are of great importance for achieving real-time requirements. In this study, several classic model compression techniques are discussed and compared by their performance on image recognition task. Additionally, an Android application able to classify images is built.