Title:Uncertainty-Driven Semantic Segmentation through Human-Machine Collaborative Learning

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.