A Novel Approach to Instance Segmentation: Integrating Mask R-CNN with Energy-Based Modeling and Points of Interest
Year: 2025
Type: conference-paper
Abstract: In this paper, a novel approach to instance segmentation is proposed by integrating Mask R-CNN with Energy-Based Modeling (EBM) and Points of Interest (PoI).Mask R-CNN, an enhancement of Faster R-CNN, is not only used for object detection but also generates segmentation masks for each object, enhancing accuracy.However, difficulties in handling small or obscured objects are often encountered in this method.To overcome these limitations, EBM is applied to provide an energy evaluation for each pixel in the image, thereby identifying high-contrast regions that need to be segmented.Meanwhile, PoI is used to determine important points in the image, optimizing the object detection and classification process.Thanks to this integration, the accuracy of segmentation is not only improved but the ability to detect objects in more complex contexts is also expanded, contributing to the development of computer vision technology.
Authors: Toan Phung Huynh, Tai Van Vo, Hiep Xuan Huynh
Institution: Can Tho University
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W4413709687