Enhancing Transparent Object Pose Estimation: A Fusion of GDR-Net and Edge Detection

2025-02-17

AI Summary

6D pose estimation of transparent objects is highly challenging due to specular reflections, refractive distortions, and background clutter. To address this, we propose an edge-driven paradigm: integrating edge preprocessing into the GDR-Net pose estimation framework. Specifically, we systematically evaluate edge detection methods—including Canny (with chromatic enhancement) and HED—applied to segmentation masks generated by YOLOX, aiming to enhance high-contrast structural features critical for robust pose estimation. To our knowledge, this is the first systematic evaluation of edge preprocessing for transparent object pose estimation under the standard BOP benchmarking protocol on the Trans6D-32K dataset. Experimental results demonstrate substantial performance gains: ADD-S accuracy improves by up to 12.7% for certain objects. These findings confirm that explicit edge features effectively mitigate perceptual ambiguities induced by transparency, offering a novel and effective direction for visual understanding of transparent objects.

Technology Category

• Computer Vision: Interpretability, Explainability, and Transparency
• Machine Learning: Transparent, Interpretable, Explainable ML
• Intelligent Robots: State Estimation

Application Category

• Security and Privacy: Data transparency and provenance
• Responsible Web: Algorithmic accountability and transparency on the web
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

Object pose estimation of transparent objects remains a challenging task in the field of robot vision due to the immense influence of lighting, background, and reflections. However, the edges of clear objects have the highest contrast, which leads to stable and prominent features. We propose a novel approach by incorporating edge detection in a pre-processing step for the tasks of object detection and object pose estimation. We conducted experiments to investigate the effect of edge detectors on transparent objects. We examine the performance of the state-of-the-art 6D object pose estimation pipeline GDR-Net and the object detector YOLOX when applying different edge detectors as pre-processing steps (i.e., Canny edge detection with and without color information, and holistically-nested edges (HED)). We evaluate the physically-based rendered dataset Trans6D-32 K of transparent objects with parameters proposed by the BOP Challenge. Our results indicate that applying edge detection as a pre-processing enhances performance for certain objects.

Problem

• Transparent object pose estimation
• Edge detection enhancement
• GDR-Net and YOLOX performance

Innovation

• Fuses GDR-Net with edge detection
• Uses Canny and HED edge detectors
• Tests on Trans6D-32K dataset