mlsanityData to 7 Oct 2026

Do Neural Networks for Segmentation Understand Insideness?

arXiv 2201.10664 ↗First posted 25 Jan 2022cs.CV cs.LG q-bio.NCcited by 0

Not on the map. Why?

Abstract (arXiv)

The insideness problem is an aspect of image segmentation that consists of determining which pixels are inside and outside a region. Deep Neural Networks (DNNs) excel in segmentation benchmarks, but it is unclear if they have the ability to solve the insideness problem as it requires evaluating long-range spatial dependencies. In this paper, the insideness problem is analysed in isolation, without texture or semantic cues, such that other aspects of segmentation do not interfere in the analysis. We demonstrate that DNNs for segmentation with few units have sufficient complexity to solve insideness for any curve. Yet, such DNNs have severe problems with learning general solutions. Only recurrent networks trained with small images learn solutions that generalize well to almost any curve. Recurrent networks can decompose the evaluation of long-range dependencies into a sequence of local operations, and learning with small images alleviates the common difficulties of training recurrent networks with a large number of unrolling steps.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 25 of its 60 references that are in the map. First posted on arXiv 25 Jan 2022.

Builds on

25 of 60 references.

References · 25

  1. 2015U-Net: Convolutional Networks for Biomedical Image Segmentationcited by 19,604
  2. 2014ImageNet Large Scale Visual Recognition Challengecited by 8,441
  3. 2017Mask R-CNNcited by 7,259
  4. 2014Fully Convolutional Networks for Semantic Segmentationcited by 4,845
  5. 2016DeepLab: Semantic Image Segmentation with Deep Convolutional Nets, Atrous Convolution, and Fully Connected CRFscited by 3,239
  6. 2018Encoder-Decoder with Atrous Separable Convolution for Semantic Image Segmentationcited by 3,210
  7. 2013Visualizing and Understanding Convolutional Networkscited by 2,278
  8. 2015SegNet: A Deep Convolutional Encoder-Decoder Architecture for Image Segmentationcited by 1,483
  9. 2015Convolutional LSTM Network: A Machine Learning Approach for Precipitation Nowcastingcited by 1,428
  10. 2015Multi-Scale Context Aggregation by Dilated Convolutionscited by 1,128
  11. 2018Path Aggregation Network for Instance Segmentationcited by 796
  12. 2017The Implicit Bias of Gradient Descent on Separable Datacited by 649
13 more
  1. 2018An Intriguing Failing of Convolutional Neural Networks and the CoordConv Solutioncited by 302
  2. 2018Recurrent Residual Convolutional Neural Network based on U-Net (R2U-Net) for Medical Image Segmentationcited by 180
  3. 2017Deep Extreme Cut: From Extreme Points to Object Segmentationcited by 127
  4. 2016Fully Convolutional Instance-aware Semantic Segmentationcited by 122
  5. 2018Learning long-range spatial dependencies with horizontal gated-recurrent unitscited by 98
  6. 2017Failures of Gradient-Based Deep Learningcited by 79
  7. 2016Memory-Efficient Backpropagation Through Timecited by 78
  8. 2017Learning to Segment Every Thingcited by 64
  9. 2017MaskLab: Instance Segmentation by Refining Object Detection with Semantic and Direction Featurescited by 54
  10. 2019Disentangling neural mechanisms for perceptual groupingcited by 28
  11. 2015ReSeg: A Recurrent Neural Network-based Model for Semantic Segmentationcited by 19
  12. 2015Iterative Instance Segmentationcited by 19
  13. 2018Cognitive Deficit of Deep Learning in Numerositycited by 4

Built on by

Not cited yet.