Uncertainty-Driven Semantic Segmentation through Human-Machine Collaborative Learning

TL;DR

This paper introduces a semi-supervised, human-in-the-loop segmentation method using cGANs that reduces manual annotation effort while maintaining high accuracy, by identifying unreliable slices and synthesizing segmentations on unlabeled data.

Contribution

It presents a novel approach combining cGANs with human-in-the-loop learning to reduce annotation needs in semantic segmentation.

Findings

Comparable to fully supervised methods in performance

Requires significantly less manual annotation

Effective in identifying unreliable slices for expert review

Abstract

Deep learning-based approaches achieve state-of-the-art performance in the majority of image segmentation benchmarks. However, training of such models requires a sizable amount of manual annotations. In order to reduce this effort, we propose a method based on conditional Generative Adversarial Network (cGAN), which addresses segmentation in a semi-supervised setup and in a human-in-the-loop fashion. More specifically, we use the discriminator to identify unreliable slices for which expert annotation is required and use the generator in the GAN to synthesize segmentations on unlabeled data for which the model is confident. The quantitative results on a conventional standard benchmark show that our method is comparable with the state-of-the-art fully supervised methods in slice-level evaluation requiring far less annotated data.