Surveillance systems integration for real time object identification using weighted bounding single neural network
In this paper, an implementation of a single neural network that classifies objects using bounding boxes and class probabilities is utilized. This features are combined with a real time surveillance system that can identify multiple targets at the same time. YOLO9000 is a contemporary tool in object...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | text |
| Published: |
Animo Repository
2017
|
| Subjects: | |
| Online Access: | https://animorepository.dlsu.edu.ph/faculty_research/2705 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | De La Salle University |
| Summary: | In this paper, an implementation of a single neural network that classifies objects using bounding boxes and class probabilities is utilized. This features are combined with a real time surveillance system that can identify multiple targets at the same time. YOLO9000 is a contemporary tool in object detection that can detect and recognize multiple targets under different categories in real-time. The system uses a multi-scale training that varies between sizes and recognizable patterns. Training of the single neural network upon detection and classification of a target varies depending upon the computer specifications. Being a classified as a simple expert system, it may less likely predict false positive results if objects are not pre-trained, but through proper intensive training and more image inputs it can predict objects in a more precise classification. This research is intended to integrate the YOLO9000 67fps concurrent monitor with surveillance hardware. © 2017 IEEE. |
|---|