TY - GEN
T1 - GAR
T2 - 2020 IEEE/CVF Winter Conference on Applications of Computer Vision, WACV 2020
AU - Li, Zheng
AU - Du, Xiaocong
AU - Cao, Yu
N1 - Funding Information:
This work was supported in part by C-BRIC, one of six centers in JUMP, a Semiconductor Research Corporation (SRC) program sponsored by DARPA. It is also partially support by National Science Foundation (CCF #1715443).
Publisher Copyright:
© 2020 IEEE.
PY - 2020/3
Y1 - 2020/3
N2 - It is well believed that object-object relations and object-scene relations inherently improve the accuracy of object detection. However, the way to efficiently model relations remains a problem. Graph Convolutional Network (GCN), an effective method to handle structured data with relations, inspires us to leverage graphs in modeling relations for objection detection tasks. In this work, we propose a novel approach, Graph Assisted Reasoning (GAR), to utilize a heterogeneous graph in modeling object-object relations and object-scene relations. GAR fuses the features from neigh-boring object nodes as well as scene nodes and produces better recognition than that produced from individual object nodes. Moreover, compared to previous approaches using Recurrent Neural Network (RNN), the light-weight and low-coupling architecture of GAR further facilitates its integration into the object detection module. Comprehensive experiments on PASCAL VOC and MS COCO datasets demonstrate the efficacy of GAR.
AB - It is well believed that object-object relations and object-scene relations inherently improve the accuracy of object detection. However, the way to efficiently model relations remains a problem. Graph Convolutional Network (GCN), an effective method to handle structured data with relations, inspires us to leverage graphs in modeling relations for objection detection tasks. In this work, we propose a novel approach, Graph Assisted Reasoning (GAR), to utilize a heterogeneous graph in modeling object-object relations and object-scene relations. GAR fuses the features from neigh-boring object nodes as well as scene nodes and produces better recognition than that produced from individual object nodes. Moreover, compared to previous approaches using Recurrent Neural Network (RNN), the light-weight and low-coupling architecture of GAR further facilitates its integration into the object detection module. Comprehensive experiments on PASCAL VOC and MS COCO datasets demonstrate the efficacy of GAR.
UR - http://www.scopus.com/inward/record.url?scp=85085530894&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85085530894&partnerID=8YFLogxK
U2 - 10.1109/WACV45572.2020.9093559
DO - 10.1109/WACV45572.2020.9093559
M3 - Conference contribution
AN - SCOPUS:85085530894
T3 - Proceedings - 2020 IEEE Winter Conference on Applications of Computer Vision, WACV 2020
SP - 1284
EP - 1293
BT - Proceedings - 2020 IEEE Winter Conference on Applications of Computer Vision, WACV 2020
PB - Institute of Electrical and Electronics Engineers Inc.
Y2 - 1 March 2020 through 5 March 2020
ER -