Related Experiment Video
Updated: Jan 14, 2026
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
3.3K
Learnable Object Queries for Few-Shot Semantic Segmentation
Summary
This study introduces a novel object-based method for few-shot semantic segmentation (FSS). The approach enhances feature extraction and utilizes prior knowledge, improving accuracy and robustness for segmenting unseen objects.
Area of Science:
- Computer Vision
- Artificial Intelligence
Background:
- Few-shot semantic segmentation (FSS) faces challenges in feature matching.
- Prototype-based methods lose detail; pixel-level methods lack robustness.
Purpose of the Study:
- To develop a target-agnostic object-based method for FSS.
- To preserve both semantic and detailed features for improved segmentation.
Main Methods:
- Introduced learnable 'object queries' for feature extraction.
- Leveraged foreground/background prior knowledge during training and inference.
- Mitigated data distribution bias using support sets and learned priors.
Main Results:
- The proposed method outperforms state-of-the-art approaches.
- Achieved superior accuracy and robustness in benchmark dataset experiments.
Conclusions:
- The object-based method effectively addresses limitations of existing FSS techniques.
- Demonstrated significant improvements in segmenting unseen objects with limited data.
