Small Sample Learning of Superpixel Classifiers for EM Segmentation- Extended Version

TL;DR

This paper introduces an interactive semi-supervised learning method for superpixel classifiers in EM segmentation, significantly reducing annotation effort while maintaining high accuracy.

Contribution

It presents an active semi-supervised approach that uses minimal labeled data and label propagation to train effective superpixel classifiers for EM segmentation.

Findings

Achieves near full-supervision accuracy with less than 20% labeled data

Reduces annotation effort in EM segmentation workflows

Demonstrates strong results across multiple datasets

Abstract

Pixel and superpixel classifiers have become essential tools for EM segmentation algorithms. Training these classifiers remains a major bottleneck primarily due to the requirement of completely annotating the dataset which is tedious, error-prone and costly. In this paper, we propose an interactive learning scheme for the superpixel classifier for EM segmentation. Our algorithm is "active semi-supervised" because it requests the labels of a small number of examples from user and applies label propagation technique to generate these queries. Using only a small set ($<20\%$) of all datapoints, the proposed algorithm consistently generates a classifier almost as accurate as that estimated from a complete groundtruth. We provide segmentation results on multiple datasets to show the strength of these classifiers.