From Transparent Labware Segmentation to Collision Avoidance: A Real-Time Edge-Aware Perception Pipeline

2026-08-05

AI Summary

This study addresses the challenge of accurately segmenting transparent laboratory glassware—a task hindered by refraction, specular reflections, and textureless surfaces—which often leads to robotic manipulation failures due to collision risks. To overcome this, the authors propose an edge-aware real-time instance segmentation framework that integrates a lightweight edge-guided attention mechanism with a parameter-free SimAM module, achieving substantially improved boundary precision with minimal model parameters. The system further enables 3D localization via multi-view mask centroid triangulation for effective obstacle avoidance. Key contributions include LabGlass-IS, the first dataset dedicated to transparent labware comprising 21 classes and 3,485 annotated images; a boundary F-score of 97.80 (an 18.93-point gain over FastSAM); inference at 7.1 ms per frame; a model size amounting to only 2.85% of competing methods; and a 93.3% success rate in real-world robotic collision avoidance.

Abstract

This paper presents an edge-aware instance segmentation framework that enables real-time robotic collision avoidance with transparent laboratory glassware using purely visual perception. Transparent vessels defy conventional segmentation due to refraction, specular reflection, and the absence of stable interior texture, yet their boundary contours remain comparatively reliable visual cues. Exploiting this observation, we augment a one-stage real-time instance segmentation backbone with a lightweight edge-detection branch, edge-guided attention fusion, and a parameter-free SimAM module, and further construct LabGlass-IS, a 3485-image, 21-category instance segmentation dataset of real laboratory glassware. The enhanced model achieves the highest Boundary F-score of 97.80 among compared methods, outperforming the YOLO-prompted FastSAM framework by 18.93 BF points. Furthermore, it maintains an inference speed of 7.1ms per frame and requires only 2.85% of the parameters of the closest accuracy competitor. Multi-view triangulation of mask centroids further provides 3D positions for conservative bounding-volume collision constraints. Real-robot trials achieve a 93.3% collision avoidance success rate, indicating the feasibility of the proposed perception-to-action pipeline for robot collision avoidance among fragile transparent objects. Our code is available at https://github.com/havishamy/TransYOLO_3D. Our video is available at https://havishamy.github.io/paper-videos/.

Problem

• transparent object segmentation
• robotic collision avoidance
• visual perception
• instance segmentation
• laboratory glassware

Innovation

• edge-aware segmentation
• transparent object perception
• real-time instance segmentation
• collision avoidance
• LabGlass-IS dataset