An Empirical Study on Multi-Domain Robust Semantic Segmentation
This work addresses the challenge of robust semantic segmentation for practical applications by improving model generalization across domains, though it is incremental in nature.
The paper tackled the problem of training a unified semantic segmentation model that performs well across multiple domains despite annotation conflicts and domain divergence, achieving a 2nd place ranking on the RVC 2022 task with a dataset only one-third the size of the first-place model.
How to effectively leverage the plentiful existing datasets to train a robust and high-performance model is of great significance for many practical applications. However, a model trained on a naive merge of different datasets tends to obtain poor performance due to annotation conflicts and domain divergence.In this paper, we attempt to train a unified model that is expected to perform well across domains on several popularity segmentation datasets.We conduct a detailed analysis of the impact on model generalization from three aspects of data augmentation, training strategies, and model capacity.Based on the analysis, we propose a robust solution that is able to improve model generalization across domains.Our solution ranks 2nd on RVC 2022 semantic segmentation task, with a dataset only 1/3 size of the 1st model used.