Inconsistency Masks: Removing the Uncertainty from Input-Pseudo-Label Pairs

25 Jan 2024  ·  Michael R. H. Vorndran, Bernhard F. Roeck ·

Efficiently generating sufficient labeled data remains a major bottleneck in deep learning, particularly for image segmentation tasks where labeling requires significant time and effort. This study tackles this issue in a resource-constrained environment, devoid of extensive datasets or pre-existing models. We introduce Inconsistency Masks (IM), a novel approach that filters uncertainty in image-pseudo-label pairs to substantially enhance segmentation quality, surpassing traditional semi-supervised learning techniques. Employing IM, we achieve strong segmentation results with as little as 10% labeled data, across four diverse datasets and it further benefits from integration with other techniques, indicating broad applicability. Notably on the ISIC 2018 dataset, three of our hybrid approaches even outperform models trained on the fully labeled dataset. We also present a detailed comparative analysis of prevalent semi-supervised learning strategies, all under uniform starting conditions, to underline our approach's effectiveness and robustness. The full code is available at: https://github.com/MichaelVorndran/InconsistencyMasks

PDF Abstract
Task Dataset Model Metric Name Metric Value Global Rank Benchmark
Semi-Supervised Semantic Segmentation Cityscapes 10% labeled IM++ (416x208, 2.7m parameters, no pretraining) Mean IoU (class) 0.428 # 1
Lesion Segmentation ISIC 2018 AIM++ (256x256, 1.5m parameters, 10% labeled data, no pretraining) Dice Score 0.85 # 11
Semi-supervised Medical Image Segmentation Lesion Segmentation on ISIC 2018 AIM++ (256x256, 1.5m parameters, 10% labeled data, no pretraining) Dice Score 0.85 # 1
Semi-Supervised Semantic Segmentation SUIM AIM+ (256x256, 2.7m parameters, 10% labeled data, no pretraining) Mean IoU (class) 0.482 # 1

Methods


No methods listed for this paper. Add relevant methods here