Flood-Filling Networks

TL;DR

Flood-filling networks offer an end-to-end trainable neural approach for 3D image segmentation, effectively handling variable object sizes and counts, and outperforming traditional multi-step pipelines in connectomic reconstruction tasks.

Contribution

This work introduces flood-filling networks, a unified neural method that directly segments images without separate boundary detection and clustering steps.

Findings

Significantly improves segmentation accuracy over existing methods.

Robustly handles images with variable object sizes and counts.

Simplifies segmentation pipeline into a single neural network.

Abstract

State-of-the-art image segmentation algorithms generally consist of at least two successive and distinct computations: a boundary detection process that uses local image information to classify image locations as boundaries between objects, followed by a pixel grouping step such as watershed or connected components that clusters pixels into segments. Prior work has varied the complexity and approach employed in these two steps, including the incorporation of multi-layer neural networks to perform boundary prediction, and the use of global optimizations during pixel clustering. We propose a unified and end-to-end trainable machine learning approach, flood-filling networks, in which a recurrent 3d convolutional network directly produces individual segments from a raw image. The proposed approach robustly segments images with an unknown and variable number of objects as well as highly variable object sizes. We demonstrate the approach on a challenging 3d image segmentation task, connectomic reconstruction from volume electron microscopy data, on which flood-filling neural networks substantially improve accuracy over other state-of-the-art methods. The proposed approach can replace complex multi-step segmentation pipelines with a single neural network that is learned end-to-end.