Stacked U-Nets: A No-Frills Approach to Natural Image Segmentation

Abstract

Many imaging tasks require global information about all pixels in an image. Conventional bottom-up classification networks globalize information by decreasing resolution; features are pooled and downsampled into a single output. But for semantic segmentation and object detection tasks, a network must provide higher-resolution pixel-level outputs. To globalize information while preserving resolution, many researchers propose the inclusion of sophisticated auxiliary blocks, but these come at the cost of a considerable increase in network size and computational cost. This paper proposes stacked u-nets (SUNets), which iteratively combine features from different resolution scales while maintaining resolution. SUNets leverage the information globalization power of u-nets in a deeper network architectures that is capable of handling the complexity of natural images. SUNets perform extremely well on semantic segmentation tasks using a small number of parameters.

arXiv commentThe code is available at https://github.com/shahsohil/sunets

References in corpus (19)

• Very Deep Convolutional Networks for Large-Scale Image Recognition
• MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications
• Rethinking Atrous Convolution for Semantic Image Segmentation
• Semantic Image Segmentation with Deep Convolutional Nets and Fully Connected CRFs
• Wide Residual Networks
• SGDR: Stochastic Gradient Descent with Warm Restarts
• ParseNet: Looking Wider to See Better
• The Cityscapes Dataset for Semantic Urban Scene Understanding
• Learning Deconvolution Network for Semantic Segmentation
• Visualizing the Loss Landscape of Neural Nets
• Stacked Hourglass Networks for Human Pose Estimation
• 3D U-Net: Learning Dense Volumetric Segmentation from Sparse Annotation
• Deformable Convolutional Networks
• Beyond Skip Connections: Top-Down Modulation for Object Detection
• Fully Connected Deep Structured Networks
• Wider or Deeper: Revisiting the ResNet Model for Visual Recognition
• Stacked Deconvolutional Network for Semantic Segmentation
• OctNetFusion: Learning Depth Fusion from Data
• An Analysis of Scale Invariance in Object Detection - SNIP

Cited by in corpus (11)

• Unsupervised Medical Image Translation Using Cycle-MedGAN
• Adversarial Inpainting of Medical Image Modalities
• A Two-step-training Deep Learning Framework for Real-time Computational Imaging without Physics Priors
• SPGNet: Semantic Prediction Guidance for Scene Parsing
• Facial UV Map Completion for Pose-invariant Face Recognition: A Novel Adversarial Approach based on Coupled Attention Residual UNets
• Accelerating Sparse Deep Neural Networks
• Unsupervised Super-Resolution of Satellite Imagery for High Fidelity Material Label Transfer
• Prostate Segmentation using 2D Bridged U-net
• UU-Nets Connecting Discriminator and Generator for Image to Image Translation
• Search Spaces for Neural Model Training
• Dense Fusion Classmate Network for Land Cover Classification