Predicting How to Distribute Work Between Algorithms and Humans to Segment an Image Batch

@article{Gurari2019PredictingHT,
  title={Predicting How to Distribute Work Between Algorithms and Humans to Segment an Image Batch},
  author={Danna Gurari and Yinan Zhao and Suyog Dutt Jain and Margrit Betke and Kristen Grauman},
  journal={International Journal of Computer Vision},
  year={2019},
  pages={1-19}
}
Foreground object segmentation is a critical step for many image analysis tasks. While automated methods can produce high-quality results, their failures disappoint users in need of practical solutions. We propose a resource allocation framework for predicting how best to allocate a fixed budget of human annotation effort in order to collect higher quality segmentations for a given batch of images and automated methods. The framework is based on a prediction module that estimates the quality of… Expand
1 Citations
Efficient Elicitation Approaches to Estimate Collective Crowd Answers
TLDR
This paper investigates eight elicitation approaches along two dimensions: annotation granularity and estimation perspective and makes it more tractable to use answer distributions in large-scale tasks such as ML training, and aims to spark future work on techniques that can efficiently estimate answer distributions. Expand

References

SHOWING 1-10 OF 61 REFERENCES
Pull the Plug? Predicting If Computers or Humans Should Segment Images
TLDR
A resource allocation framework for predicting how best to allocate a fixed budget of human annotation effort in order to collect higher quality segmentations for a given batch of images and automated methods is proposed. Expand
Predicting Sufficient Annotation Strength for Interactive Foreground Segmentation
  • S. Jain, K. Grauman
  • Computer Science
  • 2013 IEEE International Conference on Computer Vision
  • 2013
TLDR
This approach learns whether a graph cuts segmentation will succeed if initialized with a given annotation mode, based on the image's visual separability and foreground uncertainty, and shows how to select the easiest modality that will be sufficiently strong to yield high quality segmentations. Expand
Predicting Foreground Object Ambiguity and Efficiently Crowdsourcing the Segmentation(s)
TLDR
A system that automatically predicts which images are ambiguous and introduces a crowdsourcing system to achieve cost savings for collecting the diversity of all valid “ground truth” foreground object segmentations by collecting extra segmentations only when ambiguity is expected. Expand
Estimating Image Segmentation Difficulty
TLDR
A linear regression model using image features to predict segmentation difficulty level is developed, and a new feature that combines information from the log histogram of log gradient and the local binary pattern histogram is a good predictor and provides the best balance of predictive performance and model complexity. Expand
iCoseg: Interactive co-segmentation with intelligent scribble guidance
TLDR
iCoseg, an automatic recommendation system that intelligently recommends where the user should scribble next, is proposed, and users following these recommendations can achieve good quality cutouts with significantly lower time and effort than exhaustively examining all cutouts. Expand
Image Segmentation by Probabilistic Bottom-Up Aggregation and Cue Integration
TLDR
A bottom-up aggregation approach to image segmentation that takes into account intensity and texture distributions in a local area around each region and incorporates priors based on the geometry of the regions, providing a complete hierarchical segmentation of the image. Expand
Constrained parametric min-cuts for automatic object segmentation
TLDR
It is shown that this algorithm significantly outperforms the state of the art for low-level segmentation in the VOC09 segmentation dataset and achieves the same average best segmentation covering as the best performing technique to date. Expand
MILCut: A Sweeping Line Multiple Instance Learning Paradigm for Interactive Image Segmentation
TLDR
This work forms the interactive segmentation problem as a multiple instance learning (MIL) task by generating positive bags from pixels of sweeping lines within a bounding box and develops an algorithm with significant performance and efficiency gain over existing state-of-the-art systems. Expand
How to Collect Segmentations for Biomedical Images? A Benchmark Evaluating the Performance of Experts, Crowdsourced Non-experts, and Algorithms
TLDR
It is revealed that online paid crowd sourced workers without domain-specific backgrounds are reliable annotators to use as part of the laboratory protocol for segmenting biomedical images, and fusing the segmentations created by crowd sourced internet workers and algorithms yielded improved segmentation results over segmentationscreated by single crowd sourced or algorithm annotations respectively. Expand
Segmentation from a box
TLDR
A study of 14 subjects who are asked to segment a boxed target in a set of 50 real images for which they have no semantic attachment finds that the subjects perceive and trace almost the same segmentations as each other, despite the inhomogeneity of the image intensities, irregular shapes of the segmentation targets and weakness of the target boundaries. Expand
...
1
2
3
4
5
...