This is FPN-ASPP for real time semanti segmentation This is the source code for semantic segmentation. Our method achieves real-time segmentation 78 scoring in PASCAL VOC, and our model is only 100 MB in size. Our source code implementation of semantic segmentation method uses the integration of deep plab architecture and FPN ideas. Deplab architecture has the ability of multi-sensory vision, but it consumes computing time. FPN has the ability to transform low-level semantics into high-level semantics, and saves computing time. Using the advantages of both, we propose this method.
At present. This code has not been fully implemented yet, but all functions will be implemented in the near future.