.png){kind=icon}
.png){kind=icon}
Return
Integrating the edge intelligence technology into image composition: A case study
.png){kind=icon}
PRE
AI
DOI：10.1007/s12083-023-01480-2.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Image composition has important research significance and application value in image aesthetics and daily life. However, it faces problems, such as complex network structures, numerous parameters, round-trip delays, and difficult deployment. Consequently, we designed an image composition guidance system (ICGS) on mobile devices to help users capture photos with more aesthetic value through image composition. The first step was to build a lightweight object-detection model. Compressing the model by optimizing the structure and parameters reduces the size of the network model, and solves the difficult deployment problem. Second, we customized the mobile camera application development and deployed a deep learning model for this application. By reading the lighting model, we realized automatic composition guidance (ACG). Simultaneously, for scenes without objects to be detected, we designed a manual composition guidance (MCG) based on the target tracking algorithm to lock any area for composition. Furthermore, the experimental results show that the aesthetic scores of the guided photos improve, which is more in line with public aesthetics. In addition, the application's real-time performance, stability, and response time have also reached high standards.
Keywords:
Image composition
Edge intelligence
Object detection
Model compression
TensorFlow lite
Journal
IF:
2.6
Papers:
2.2K
Citations:
2.9K