Skip to main content

Lund University Publications

LUND UNIVERSITY LIBRARIES

Deep Reinforcement Learning of Region Proposal Networks for Object Detection

Pirinen, Aleksis LU and Sminchisescu, Cristian LU (2018) 31st Meeting of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2018 p.6945-6954
Abstract

We propose drl-RPN, a deep reinforcement learning-based visual recognition model consisting of a sequential region proposal network (RPN) and an object detector. In contrast to typical RPNs, where candidate object regions (RoIs) are selected greedily via class-agnostic NMS, drl-RPN optimizes an objective closer to the final detection task. This is achieved by replacing the greedy RoI selection process with a sequential attention mechanism which is trained via deep reinforcement learning (RL). Our model is capable of accumulating class-specific evidence over time, potentially affecting subsequent proposals and classification scores, and we show that such context integration significantly boosts detection accuracy. Moreover, drl-RPN... (More)

We propose drl-RPN, a deep reinforcement learning-based visual recognition model consisting of a sequential region proposal network (RPN) and an object detector. In contrast to typical RPNs, where candidate object regions (RoIs) are selected greedily via class-agnostic NMS, drl-RPN optimizes an objective closer to the final detection task. This is achieved by replacing the greedy RoI selection process with a sequential attention mechanism which is trained via deep reinforcement learning (RL). Our model is capable of accumulating class-specific evidence over time, potentially affecting subsequent proposals and classification scores, and we show that such context integration significantly boosts detection accuracy. Moreover, drl-RPN automatically decides when to stop the search process and has the benefit of being able to jointly learn the parameters of the policy and the detector, both represented as deep networks. Our model can further learn to search over a wide range of exploration-accuracy trade-offs making it possible to specify or adapt the exploration extent at test time. The resulting search trajectories are image- and category-dependent, yet rely only on a single policy over all object categories. Results on the MS COCO and PASCAL VOC challenges show that our approach outperforms established, typical state-of-the-art object detection pipelines.

(Less)
Please use this url to cite or link to this publication:
author
and
organization
publishing date
type
Chapter in Book/Report/Conference proceeding
publication status
published
subject
host publication
Proceedings - 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2018
article number
8578824
pages
10 pages
publisher
IEEE Computer Society
conference name
31st Meeting of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2018
conference location
Salt Lake City, United States
conference dates
2018-06-18 - 2018-06-22
external identifiers
  • scopus:85062828217
ISBN
9781538664209
DOI
10.1109/CVPR.2018.00726
language
English
LU publication?
yes
id
5610ef4e-3535-43a3-b501-5fda2996c757
date added to LUP
2019-04-01 09:52:21
date last changed
2022-05-03 18:57:08
@inproceedings{5610ef4e-3535-43a3-b501-5fda2996c757,
  abstract     = {{<p>We propose drl-RPN, a deep reinforcement learning-based visual recognition model consisting of a sequential region proposal network (RPN) and an object detector. In contrast to typical RPNs, where candidate object regions (RoIs) are selected greedily via class-agnostic NMS, drl-RPN optimizes an objective closer to the final detection task. This is achieved by replacing the greedy RoI selection process with a sequential attention mechanism which is trained via deep reinforcement learning (RL). Our model is capable of accumulating class-specific evidence over time, potentially affecting subsequent proposals and classification scores, and we show that such context integration significantly boosts detection accuracy. Moreover, drl-RPN automatically decides when to stop the search process and has the benefit of being able to jointly learn the parameters of the policy and the detector, both represented as deep networks. Our model can further learn to search over a wide range of exploration-accuracy trade-offs making it possible to specify or adapt the exploration extent at test time. The resulting search trajectories are image- and category-dependent, yet rely only on a single policy over all object categories. Results on the MS COCO and PASCAL VOC challenges show that our approach outperforms established, typical state-of-the-art object detection pipelines.</p>}},
  author       = {{Pirinen, Aleksis and Sminchisescu, Cristian}},
  booktitle    = {{Proceedings - 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2018}},
  isbn         = {{9781538664209}},
  language     = {{eng}},
  month        = {{12}},
  pages        = {{6945--6954}},
  publisher    = {{IEEE Computer Society}},
  title        = {{Deep Reinforcement Learning of Region Proposal Networks for Object Detection}},
  url          = {{http://dx.doi.org/10.1109/CVPR.2018.00726}},
  doi          = {{10.1109/CVPR.2018.00726}},
  year         = {{2018}},
}