Pith. Sign in

RECORD

Fast Disparity Estimation using Dense Networks

PaperPDF from arXiv

data files

Paper breakdownThe authors' claims, conditions and measurements, as JSON Signed receiptProof this review has not been altered
X Bluesky LinkedIn Reddit Hacker News
Rowel Atienza

Computer Vision and Pattern Recognition · Submitted 19 May 2018

abstract compiled · 2026-09-23 · no language-model review

The authors' main claim Experimental results show that DenseMapNet accuracy is comparable with other significantly bigger CNN-based methods.

how it is stated · asserted 1.00 read the record →

Comment (0) Read on arXiv ↗
Cite ▾
arXiv ID1805.07499 (version 1)
Pith ID5EFUXMCL, Pith's permanent record of this paper and its review
CitationRowel Atienza. "Fast Disparity Estimation using Dense Networks." arXiv:1805.07499 (2018). https://pith.science/paper/5EFUXMCL
Claim paper

The compiled record

The paper's own sentences, arranged. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. Numbers in the tables are the authors', checked by code. No sentence in this record was written by a model.

Open the compiled record

Compiled by classifier-1.13.0 via TypeSafe API, question bank abstract-p1-20260923, 2026-09-23: 1 request, 6,046 input tokens.

What the abstract itself states, as probabilities: claims first 0.04 · states limitation 0.13 · human participants 0.02 · controlled comparison 0.42 · randomized assignment 0.01 · states quantitative result 0.98 · compares with prior results 0.94 · mentions code or data availability 0.03.

The abstract, sentence by sentence

arXiv listing abstract v1. Results and the main claim are underlined; .

Disparity estimation is a difficult problem in stereo vision because the correspondence technique fails in images with textureless and repetitive regions. Recent body of work using deep convolutional neural networks (CNN) overcomes this problem with semantics. Most CNN implementations use an autoencoder method; stereo images are encoded, merged and finally decoded to predict the disparity map. In this paper, we present a CNN implementation inspired by dense networks to reduce the number of parameters. Furthermore, our approach takes into account semantic reasoning in disparity estimation. Our proposed network, called DenseMapNet, is compact, fast and can be trained end-to-end. DenseMapNet requires 290k parameters only and runs at 30Hz or faster on color stereo images in full resolution. Experimental results show that DenseMapNet accuracy is comparable with other significantly bigger CNN-based methods.

background contribution method result interpretation main claim

What the classifier reports for each sentence. The number is the probability it gave the label.
#sentencerolethe authors' own workstates the main conclusionstrength
s0 Disparity estimation is a difficult problem in stereo vision because the correspondence technique fails in images with textureless and repetitive regions. background 1.00 0.06 0.02 asserted 1.00
s1 Recent body of work using deep convolutional neural networks (CNN) overcomes this problem with semantics. background 1.00 0.03 0.03 asserted 1.00
s2 Most CNN implementations use an autoencoder method; stereo images are encoded, merged and finally decoded to predict the disparity map. background 0.99 0.05 0.03 asserted 1.00
s3 In this paper, we present a CNN implementation inspired by dense networks to reduce the number of parameters. contribution 0.90 0.86 0.07 asserted 0.99
s4 Furthermore, our approach takes into account semantic reasoning in disparity estimation. contribution 0.72 0.85 0.07 asserted 1.00
s5 Our proposed network, called DenseMapNet, is compact, fast and can be trained end-to-end. contribution 0.89 0.86 0.29 asserted 1.00
s6 DenseMapNet requires 290k parameters only and runs at 30Hz or faster on color stereo images in full resolution. result 0.85 0.94 0.20 asserted 1.00
s7 Experimental results show that DenseMapNet accuracy is comparable with other significantly bigger CNN-based methods. result 1.00 0.95 0.58 asserted 1.00

What kind of paper the classifier reports this is: method or tool 1.00 · evidence from observation or experiment 0.98.

What was examined, and what was not

sources read
  • arXiv listing abstract v1
not examined
  • full text
  • figures and tables
  • references and citing sentences
  • relations to other papers
levels run
  • P abstract
not run
  • S sections
  • T body sentences
  • C citations
  • R cross-paper relations

Every quoted sentence on this record is the authors'. Every label is from Pith's closed vocabulary, with the probability shown beside it. Every number in the measurements table is the authors', checked by code. Record as JSON

The paper
Original abstract

Disparity estimation is a difficult problem in stereo vision because the correspondence technique fails in images with textureless and repetitive regions. Recent body of work using deep convolutional neural networks (CNN) overcomes this problem with semantics. Most CNN implementations use an autoencoder method; stereo images are encoded, merged and finally decoded to predict the disparity map. In this paper, we present a CNN implementation inspired by dense networks to reduce the number of parameters. Furthermore, our approach takes into account semantic reasoning in disparity estimation. Our proposed network, called DenseMapNet, is compact, fast and can be trained end-to-end. DenseMapNet requires 290k parameters only and runs at 30Hz or faster on color stereo images in full resolution. Experimental results show that DenseMapNet accuracy is comparable with other significantly bigger CNN-based methods.

Human review

How Pith checks papers → · Full audit of this paper →

Discussion

Know something about this paper? Leave a comment, or highlight any passage above to comment on it.