Tackling the Problem of Limited Data and Annotations in Semantic Segmentation

In this work, the case of semantic segmentation on a small image dataset (simulated by 1000 randomly selected images from PASCAL VOC 2012), where only weak supervision signals (scribbles from user interaction) are available is studied. Especially, to tackle the problem of limited data annotations in image segmentation, transferring different pre-trained models and CRF based methods are applied to enhance the segmentation performance. To this end, RotNet, DeeperCluster, and SemiW moreover, for the case of training on the full PASCAL VOC 2012 training data, this pre-training approach increases the mIoU results by almost 4%. On the other hand, dense CRF is shown to be very effective as well, enhancing the results both as a loss regularization technique in weakly supervised training and as a post-processing tool.

• Publication year

  2020

• Venue

  arXiv.org

• Publication date

  2020-07-14

• Fields of study

  Computer Science

• Identifiers
  arXiv 2007.07357
• External record

  Open on Semantic Scholar

• Source metadata

  Semantic Scholar

• No claims are published for this paper.

• No concepts are published for this paper.

• Revisiting Self-Supervised Visual Representation Learning

  2019cited by this paper
• Unsupervised Pre-Training of Image Features on Non-Curated Data

  2019influential reference
• A survey on semi-supervised learning

  2019cited by this paper
• Billion-scale semi-supervised learning for image classification

  2019influential reference
• Self-Supervised Representation Learning by Rotation Feature Decoupling

  2019cited by this paper
• A brief introduction to weakly supervised learning

  2018cited by this paper
• Exploring the Limits of Weakly Supervised Pretraining

  2018cited by this paper
• Deep Clustering for Unsupervised Learning of Visual Features

  2018cited by this paper
• Unsupervised Representation Learning by Predicting Image Rotations

  2018cited by this paper
• Boosting Self-Supervised Learning via Knowledge Transfer

  2018cited by this paper
• Normalized Cut Loss for Weakly-Supervised CNN Segmentation

  2018cited by this paper
• On Regularized Losses for Weakly-supervised CNN Segmentation

  2018influential reference
• Taskonomy: Disentangling Task Transfer Learning

  2018cited by this paper
• Representation Learning with Contrastive Predictive Coding

  2018cited by this paper
• Multi-task Self-Supervised Visual Learning

  2017cited by this paper
• Semi-Supervised Learning with Generative Adversarial Networks

  2016cited by this paper
• DeepCut: Object Segmentation From Bounding Box Annotations Using Convolutional Neural Networks

  2016cited by this paper
• Seed, Expand and Constrain: Three Principles for Weakly-Supervised Image Segmentation

  2016cited by this paper
• The Cityscapes Dataset for Semantic Urban Scene Understanding

  2016cited by this paper
• Unsupervised Learning of Visual Representations by Solving Jigsaw Puzzles

  2016cited by this paper
• DeepLab: Semantic Image Segmentation with Deep Convolutional Nets, Atrous Convolution, and Fully Connected CRFs

  2016influential reference
• ScribbleSup: Scribble-Supervised Convolutional Networks for Semantic Segmentation

  2016influential reference
• Weakly-and Semi-Supervised Learning of a Deep Convolutional Network for Semantic Image Segmentation

  2015cited by this paper
• Modeling local and global deformations in Deep Learning: Epitomic convolution, Multiple Instance Learning, and sliding window detection

  2015cited by this paper
• Conditional Random Fields as Recurrent Neural Networks

  2015cited by this paper
• Constrained Convolutional Neural Networks for Weakly Supervised Segmentation

  2015cited by this paper
• Unsupervised Visual Representation Learning by Context Prediction

  2015cited by this paper
• Weakly Supervised Deep Detection Networks

  2015cited by this paper
• The Fast Bilateral Solver

  2015cited by this paper
• Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks

  2015cited by this paper
• Object Detection in Optical Remote Sensing Images Based on Weakly Supervised Learning and High-Level Feature Learning

  2015cited by this paper
• Microsoft COCO: Common Objects in Context

  2014cited by this paper
• The Pascal Visual Object Classes Challenge: A Retrospective

  2014cited by this paper
• Fully convolutional networks for semantic segmentation

  2014influential reference
• Semi-supervised Learning with Deep Generative Models

  2014cited by this paper
• Efficient Inference in Fully Connected CRFs with Gaussian Edge Potentials

  2011cited by this paper
• ImageNet: A large-scale hierarchical image database

  2009cited by this paper
• Class segmentation and object localization with superpixel neighborhoods

  2009cited by this paper
• Introduction to Semi-Supervised Learning

  2009cited by this paper
• A Fast Approximation of the Bilateral Filter Using a Signal Processing Approach

  2006cited by this paper
• Snakes: Active contour models

  2004cited by this paper
• "GrabCut": interactive foreground extraction using iterated graph cuts

  2004cited by this paper
• Multi-label Image Segmentation for Medical Applications Based on Graph-Theoretic Electrical Potentials

  2004cited by this paper
• Computing geodesics and minimal surfaces via graph cuts

  2003cited by this paper
• Image segmentation by histogram thresholding using fuzzy sets

  2002cited by this paper
• Interactive graph cuts for optimal boundary & region segmentation of objects in N-D images

  2001cited by this paper
• Toboggan-based intelligent scissors with a four-parameter edge model

  1999cited by this paper
• Seeded Region Growing

  1994cited by this paper
• A real-time algorithm for signal analysis with the help of the wavelet transform

  1989cited by this paper

• No citing papers are available for this paper.