Abstract: Semantic segmentation is challenging as it requires both object-level information and pixel-level accuracy. Recently, FCN-based systems have great improvement in this area. Unlike classification networks, combining features of different layers plays an important role in these dense prediction models, as these features contain information of different levels. A number of models have been proposed to show how to use these features. However, what is the best architecture to make use of features of different layers is still a question. This paper proposes a module, called mixed context network, and shows that the presented system outperforms most existing semantic segmentation systems by making use of this module. This model performs well in the court scene.

Key words: semantic segmentation, Fully Convolutional Networks（FCN）, mixed context network, court scene

摘要： 语义分割需要兼顾目标级的高级语义信息和像素级的准确性，所以非常有挑战性。最近，基于全卷积网络的系统在这个领域取得了很大的进展。和分类网络不同，在这些密集预测模型中，综合来自不同层的特征尤为重要，因为这些特征包含着不同级别的信息。什么样的网络结构才能最好地利用这些特征仍然是一个开放的问题。提出了一种混合上下文网络的模块，加入该模块的语义分割系统表现出了非常优越的性能，在庭审场景下亦表现良好。

关键词: 语义分割, 全卷积网络, 混合上下文网络, 庭审场景