Delving into Crispness: Guided Label Refinement for Crisp Edge Detection

TL;DR

This paper identifies noisy human labels as the main cause of thick edges in learned edge detection and proposes a Canny-guided label refinement method to produce crisper edge labels, significantly improving detection quality.

Contribution

It introduces a Canny-guided label refinement technique that enhances label quality for training more precise edge detectors, emphasizing label quality over model complexity.

Findings

Refined labels lead to a significant increase in edge crispness from 17.4% to 30.6%.

Training with refined edges improves ODS and OIS by over 12% on the Multicue dataset.

Crisp edge detection benefits downstream tasks like optical flow and image segmentation.

Abstract

Learning-based edge detection usually suffers from predicting thick edges. Through extensive quantitative study with a new edge crispness measure, we find that noisy human-labeled edges are the main cause of thick predictions. Based on this observation, we advocate that more attention should be paid on label quality than on model design to achieve crisp edge detection. To this end, we propose an effective Canny-guided refinement of human-labeled edges whose result can be used to train crisp edge detectors. Essentially, it seeks for a subset of over-detected Canny edges that best align human labels. We show that several existing edge detectors can be turned into a crisp edge detector through training on our refined edge maps. Experiments demonstrate that deep models trained with refined edges achieve significant performance boost of crispness from 17.4% to 30.6%. With the PiDiNet backbone, our method improves ODS and OIS by 12.2% and 12.6% on the Multicue dataset, respectively, without relying on non-maximal suppression. We further conduct experiments and show the superiority of our crisp edge detection for optical flow estimation and image segmentation.