Войти
аналитика - Instance Segmentation
CenterDisks: Real-time Instance Segmentation with Disk Covering
Using disk covering for real-time instance segmentation achieves state-of-the-art results with efficient resource utilization.
Self-Balanced R-CNN: An Evolved Instance Segmentation Architecture with Improved Imbalance Handling
The Self-Balanced R-CNN (SBR-CNN) architecture addresses key imbalance issues in two-stage instance segmentation models, including IoU distribution imbalance and feature-level imbalance, through novel loop mechanisms and an improved RoI extraction layer.
Synthetic Instance Segmentation from Semantic Image Segmentation Masks: A Novel Approach for Efficient and Accurate Object Recognition
This paper introduces SISeg, a novel instance segmentation method that leverages existing semantic segmentation models to achieve accurate object recognition without requiring instance-level annotations, thereby improving efficiency and reducing annotation costs.