Scene Parsing with Global Context Embedding
This addresses scene parsing accuracy for computer vision applications, though it appears incremental as it builds on existing segmentation networks with added context modules.
The paper tackles scene parsing by incorporating global context information through learned feature representations and priors, which reduces false positives incompatible with global context. Experiments on MIT ADE20K and PASCAL Context datasets show the method performs favorably against existing methods.
We present a scene parsing method that utilizes global context information based on both the parametric and non- parametric models. Compared to previous methods that only exploit the local relationship between objects, we train a context network based on scene similarities to generate feature representations for global contexts. In addition, these learned features are utilized to generate global and spatial priors for explicit classes inference. We then design modules to embed the feature representations and the priors into the segmentation network as additional global context cues. We show that the proposed method can eliminate false positives that are not compatible with the global context representations. Experiments on both the MIT ADE20K and PASCAL Context datasets show that the proposed method performs favorably against existing methods.