Search papers, labs, and topics across Lattice.
This paper introduces Difficulty-Aware Sample Allocation (DASA), a novel framework that enhances data augmentation in semantic segmentation by assigning stronger transformations to samples deemed more difficult based on a multi-factorial difficulty score. By integrating factors such as prediction ambiguity, training loss, class rarity, and boundary complexity, DASA effectively tailors augmentation strength to individual samples during training. Experimental results demonstrate that DASA significantly improves segmentation performance, achieving a mean Intersection over Union (mIoU) increase from 0.633 to 0.740 for DeepLabV3 on the Oxford-IIIT Pet dataset and outperforming standard approaches on binary Pascal VOC.
Multi-factor difficulty estimation boosts segmentation performance, achieving state-of-the-art results across multiple architectures.
Data augmentation is a standard component of modern semantic segmentation pipelines, but most augmentation techniques allocate transformations uniformly across training samples or adapt to a single difficulty signal such as loss. This ignores the fact that segmentation difficulty is multi-factorial, since ambiguous predictions, persistent optimization errors, rare classes, and complex object boundaries can each make a sample informative in different ways. This paper introduces Difficulty-Aware Sample Allocation (DASA), an architecture-agnostic framework that assigns stronger augmentation to samples estimated to be more difficult. DASA combines prediction ambiguity, training loss, class rarity, and boundary complexity into a normalized difficulty score, then maps that score to sample-specific augmentation strength during iterative training. Experiments on Oxford-IIIT Pet and binary Pascal VOC segmentation with U-Net, DeepLabV3, and SegFormer-B0 show that DASA improves over standard training and is competitive with or stronger than single-signal adaptive baselines. On Oxford-IIIT Pet, DASA improves DeepLabV3 from 0.633 to 0.740 mIoU. On binary Pascal VOC, DASA obtains the best foreground IoU for all three evaluated architectures. These results attest to the value of multi-factor difficulty estimation as a practical mechanism for directing augmentation where it is most useful.
