Predicting Signed Distance Functions for Visual Instance Segmentation

2026-08-13

AI Summary

This work addresses the challenge of accurately modeling highly variable or elongated objects—such as ropes—in instance segmentation, where conventional anchor-based methods often fail to capture complex shapes. To overcome this limitation, the authors propose a novel anchor-free paradigm that leverages neural networks to predict, at each pixel, multi-directional distances to the nearest object boundary. These predictions are aggregated to approximate a signed distance function (SDF), from which a foreground mask is obtained via thresholding. By performing pixel-wise directional distance regression, the method flexibly represents arbitrary object geometries without relying on predefined anchors. Evaluated on the COCO dataset, the approach achieves superior segmentation IoU compared to state-of-the-art methods like YOLACT, demonstrating significantly enhanced adaptability to irregular object shapes.

Abstract

Visual instance segmentation is a challenging problem and becomes even more difficult if objects of interest varies unconstrained in shape. Some objects are well described by a rectangle, however, this is hardly always the case. Consider for instance long, slender objects such as ropes. Anchor-based approaches classify predefined bounding boxes as either negative or positive and thus provide a limited set of shapes that can be handled. Defining anchor-boxes that fit well to all possible shapes leads to an infeasible number of prior boxes. We explore a different approach and propose to train a neural network to compute distance maps along different directions. The network is trained at each pixel to predict the distance to the closest object contour in a given direction. By pooling the distance maps we obtain an approximation to the signed distance function (SDF). The SDF may then be thresholded in order to obtain a foreground-background segmentation. We compare this segmentation to foreground segmentations obtained from the state-of-the-art instance segmentation method YOLACT. On the COCO dataset, our segmentation yields a higher performance in terms of foreground intersection over union (IoU). However, while the distance maps contain information on the individual instances, it is not straightforward to map them to the full instance segmentation. We still believe that this idea is a promising research direction for instance segmentation, as it better captures the different shapes found in the real world.

Problem

• visual instance segmentation
• signed distance function
• shape variability
• anchor-based methods
• object contour

Innovation

• Signed Distance Function
• Instance Segmentation
• Distance Map
• Anchor-free
• Shape Representation