Generating Accurate Candidate Windows by Effective Receptive Field

Baojun ZHAO  Boya ZHAO  Linbo TANG  Baoxian WANG  

Publication
IEICE TRANSACTIONS on Fundamentals of Electronics, Communications and Computer Sciences   Vol.E102-A   No.12   pp.1925-1927
Publication Date: 2019/12/01
Online ISSN: 1745-1337
DOI: 10.1587/transfun.E102.A.1925
Type of Manuscript: Special Section LETTER (Special Section on Smart Multimedia & Communication Systems)
Category: Image
Keyword: 
convolutional neural networks,  object detection,  feature pyramid,  anchor generation,  

Full Text: PDF(713.7KB)>>
Buy this Article



Summary: 
Towards involving the convolutional neural networks into the object detection field, many computer vision tasks have achieved favorable successes. In order to adapt targets with various scales, deep feature pyramid is widely used, since the traditional object detection methods detect different objects in Gaussian image pyramid. However, due to the mismatching between the anchors and the feature distributions of targets, the accurate detection for targets with various scales is still a challenge. Considering the differences between the theoretical receptive field and effective receptive field, we propose a novel anchor generation method, which takes the effective receptive field as the standard. The proposed method is evaluated on the PASCAL VOC dataset and shows the favorable results.