Select, Attend, and Transfer: Light, Learnable Skip Connections
Saeid Asgari Taghanaki, Aicha Bentaieb, Anmol Sharma, S. Kevin Zhou,, Yefeng Zheng, Bogdan Georgescu, Puneet Sharma, Sasa Grbic, Zhoubing Xu, Dorin, Comaniciu, Ghassan Hamarneh

TL;DR
This paper introduces light, learnable skip connections for segmentation networks that select discriminative features and regions, reducing memory and parameters while improving accuracy across multiple datasets.
Contribution
The paper proposes a novel learnable skip connection method that enhances segmentation performance and efficiency by focusing on discriminative features and regions.
Findings
Outperforms traditional skip connections in accuracy
Reduces memory and parameter usage
Effective on 2D and volumetric datasets
Abstract
Skip connections in deep networks have improved both segmentation and classification performance by facilitating the training of deeper network architectures, and reducing the risks for vanishing gradients. They equip encoder-decoder-like networks with richer feature representations, but at the cost of higher memory usage, computation, and possibly resulting in transferring non-discriminative feature maps. In this paper, we focus on improving skip connections used in segmentation networks (e.g., U-Net, V-Net, and The One Hundred Layers Tiramisu (DensNet) architectures). We propose light, learnable skip connections which learn to first select the most discriminative channels and then attend to the most discriminative regions of the selected feature maps. The output of the proposed skip connections is a unique feature map which not only reduces the memory usage and network parameters to a…
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
MethodsConcatenated Skip Connection · *Communicated@Fast*How Do I Communicate to Expedia? · Max Pooling · Convolution · U-Net
