A Method for Object Segmentation in Low-Resolution Images Based on Mask R-CNN
- Paper number
IAC-19,B1,IP,17,x50937
- Author
Ms. Ting Da, China, Xi'an Microelectronics Technology Institute, China Aerospace Science and Technology Corporation (CASC)
- Coauthor
Mr. Yang Liang, China
- Coauthor
Dr. Liang Yang, China, Xi'an Microelectronics Technology Institute, China Aerospace Science and Technology Corporation (CASC)
- Year
2019
- Abstract
A novel method for object segmentation in low-resolution images based on Mask R-CNN is proposed. Although the existing methods have made remarkable progress in high-resolution images and large-sized objects, the performance on low-resolution images or small objects is far from satisfactory. Since it is difficult to obtain the high-resolution images in many cases, we aim to pick objects in low-resolution images with higher accuracy using deep learning. A new model, which is designed to combine the bicubic interpolation and Mask R-CNN, is developed to improve the picking performance in cityscape images. The experimental results show that the proposed method can get a better performance compared to existing algorithms.
- Abstract document
- Manuscript document
(absent)