| Title |
Implementation and Performance Analysis of Improved Annotation System for Efficient Object Detection |
| Authors |
류병선(Byeongseon Ryu) ; 강문식(Moonsik Kang) |
| DOI |
https://doi.org/10.5573/ieie.2024.61.4.31 |
| Keywords |
Object detection; Annotation system; Class processing; Learning data; Image segmentation |
| Abstract |
In this paper, we propose an efficient annotation system to overcome the limitations of existing annotation schemes and systems. The new annotation system provides consistent class processing to increase data reliability. Additionally, using image segmentation, training data of various sizes can be effectively utilized without adding annotation work. As a result of comparing before and after applying the designed module processing function, it was confirmed that the more consistency is maintained, the more the reliability of the data is improved, and the proposed object detection model shows higher performance. By applying the proposed system to practical use, error problems that may occur during annotation work have been resolved, and reliable and efficient object detection is achieved as well as the construction of a precise object detection model through more flexible use of learning data. |