BROWSE
Detail
Conference Paper
An Optimization Tool for Local Customized Object Detector in Edge Devices
Cited 0 time in 
Share 
Share
- Authors
- Jang Woon Baek, Yun Won Choi, Jinhong Kim, Joon-Goo Lee
- Issue Date
- 2024-02
- Citation
- International Conference on Artificial Intelligence in Information and Communication (ICAIIC) 2024, pp.702-704
- Language
- English
- Type
- Conference Paper
- Abstract
- In this paper, we propose an optimization tool for local customized object detector which has a lightweight deep learning model and operates on the inexpensive and low-performance edge devices. We focused on the edge devices which provide video analysis services on a CCTV camera that monitors a specific area. The pre-trained object detection model needs customizing processes such as re-learning depending on the camera location in order to reduce false alarm and miss detection. That requires additional time and energy to create a training DB from the video of the CCTV camera and to train the detection model using the created DB. In order to reduce this effort, we developed a web-based optimization tool for a local customized object detector. The proposed tool provides the automatic DB generation and re-learning functions with user-friendly interfaces. We can see that the proposed tool is very simple and efficient which only requires a video file from the local camera and the pre-trained detection model, while automatic DB generation and re-learning processes are done internally.
- KSP Keywords
- Analysis services, CCTV Camera, Detection model, Edge devices, Learning Process, Specific area, Time and Energy, User-friendly, Video file, deep learning(DL), false alarm and miss detection