DAG-Recurrent Neural Networks For Scene Labeling
arXiv:1509.00552
Abstract
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.
References in corpus (4)
Cited by in corpus (11)
- A Review on Deep Learning Techniques Applied to Semantic Segmentation
- Recent Advances in Convolutional Neural Networks
- Deep Learning Algorithms with Applications to Video Analytics for A Smart City: A Survey
- A Siamese Long Short-Term Memory Architecture for Human Re-Identification
- Ithemal: Accurate, Portable and Fast Basic Block Throughput Estimation using Deep Neural Networks
- Improving Fully Convolution Network for Semantic Segmentation
- SANet: Structure-Aware Network for Visual Tracking
- Multi-Path Feedback Recurrent Neural Network for Scene Parsing
- Neuron-level Selective Context Aggregation for Scene Segmentation
- Scene Labeling using Gated Recurrent Units with Explicit Long Range Conditioning
- Scene Parsing via Dense Recurrent Neural Networks with Attentional Selection