Attention-refined two-branch networks for real-time semantic segmentation

HIGHLIGHTS

• What: In this paper an efficient Attention Refined Two-Branch Real-Time Semantic Segmentation Network (ARTRNet) is designed to alleviate the above challenges. The authors propose a lightweight dense connectivity context refinement branch consisting of a novel DownSampling Module (DSM) and a lightweight dense feature module, which achieves high efficiency with reduced computational cost and model size. The authors propose an Attention Refinement Module (ARM). to compute the attention vector of each feature map as a way to highlight features. The authors compare the final accuracy and speed (FPS) results with other algorithms across various benchmarks.
 



If you want to have access to all the content you need to log in!

Thanks :)

Username or Email

Password

Remember me

Lost your password?

If you don't have an account, you can create one here.