DenseNet: Implementing Efficient ConvNet Descriptor Pyramids

Forrest Iandola, Matt Moskewicz, Sergey Karayev, Ross Girshick, Trevor Darrell +1 more
4/7/2014
cs.CV

Official code: forresti/caffe (5 stars, last commit 2015-05-24).

Abstract

Convolutional Neural Networks (CNNs) can provide accurate object classification. They can be extended to perform object detection by iterating over dense or selected proposed object regions. However, the runtime of such detectors scales as the total number and/or area of regions to examine per image, and training such detectors may be prohibitively slow. However, for some CNN classifier topologies, it is possible to share significant work among overlapping regions to be classified. This paper presents DenseNet, an open source system that computes dense, multiscale features from the convolutional layers of a CNN based object classifier. Future work will involve training efficient object detectors with DenseNet feature descriptors.

View on arXivView PDF

Code Implementations(1)

forresti/caffeOfficial100%

Caffe

52C++, CSSJan 15, 201411 years agoNOASSERTION

Similar Papers

R-FCN: Object Detection via Region-based Fully Convolutional Networks

86% match1 repo1.2k2016

PVANET: Deep but Lightweight Neural Networks for Real-time Object Detection

86% match1 repo6492016

Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks

86% match2 repos8.3k2015

DSOD: Learning Deeply Supervised Object Detectors from Scratch

86% match1 repo7102017

MuRAL: Multi-Scale Region-based Active Learning for Object Detection

85% match2023

Fast R-CNN

85% match1 repo3.5k2015

Cite this paper

@article{iandola2014densenet,
  title  = {DenseNet: Implementing Efficient ConvNet Descriptor Pyramids},
  author = {Forrest Iandola and Matt Moskewicz and Sergey Karayev and Ross Girshick and Trevor Darrell and Kurt Keutzer},
  year   = {2014},
  eprint = {1404.1869},
  archivePrefix = {arXiv},
  url    = {https://arxiv.org/abs/1404.1869}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/1404.1869.svg)](https://codeofpaper.com/paper/1404.1869)

Discussion