DAG-Recurrent Neural Networks For Scene Labeling
This addresses the issue of ineffective long-range context encoding in scene labeling for computer vision applications, representing an incremental improvement with novel method integration.
The paper tackled the problem of encoding long-range contextual information in image labeling by proposing directed acyclic graph recurrent neural networks (DAG-RNNs) to model semantic dependencies, achieving new state-of-the-art results on SiftFlow, CamVid, and Barcelona benchmarks.
In image labeling, local representations for image units are usually generated from their surrounding image patches, thus long-range contextual information is not effectively encoded. In this paper, we introduce recurrent neural networks (RNNs) to address this issue. Specifically, directed acyclic graph RNNs (DAG-RNNs) are proposed to process DAG-structured images, which enables the network to model long-range semantic dependencies among image units. Our DAG-RNNs are capable of tremendously enhancing the discriminative power of local representations, which significantly benefits the local classification. Meanwhile, we propose a novel class weighting function that attends to rare classes, which phenomenally boosts the recognition accuracy for non-frequent classes. Integrating with convolution and deconvolution layers, our DAG-RNNs achieve new state-of-the-art results on the challenging SiftFlow, CamVid and Barcelona benchmarks.