This is the authors' abstract. We don't add key points for this paper.
We propose a novel deep architecture, SegNet, for semantic pixel wise image. SegNet has several attractive properties; (i) it only requires evaluation of a fully learnt function to obtain smooth label, (ii) with increasing depth, a larger context is considered for labelling which improves accuracy, and (iii) it is easy to visualise the of feature activation(s) in the pixel label space at any depth. SegNet composed of a stack of encoders followed by a corresponding decoder stack feeds into a soft-max classification layer. The decoders help map low feature maps at the output of the encoder stack to full input image feature maps. This addresses an important drawback of recent deep learning which have adopted networks designed for object categorization for wise labelling. These methods lack a mechanism to map deep layer feature to input dimensions. They resort to ad hoc methods to upsample features,≠.g. by replication. This results in noisy predictions and also restricts the of pooling layers in order to avoid too much upsampling and thus reduces context. SegNet overcomes these problems by learning to map encoder to image pixel labels. We test the performance of SegNet on outdoor RGB from CamVid, KITTI and indoor scenes from the NYU dataset. Our results that SegNet achieves state-of-the-art performance even without use of cues such as depth, video frames or post-processing with CRF models.
Answers are grounded in this paper and cite their sources.
No takes yet. Share an insight, caveat, or question.
Badrinarayanan et al. (2015) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: