ABOUT JoVE

Overview Leadership Blog JoVE Help Center

AUTHORS

Publishing Process Editorial Board Scope & Policies Peer Review FAQ Submit

LIBRARIANS

Testimonials Subscriptions Access Resources Library Advisory Board FAQ

RESEARCH

JoVE Journal Methods Collections JoVE Encyclopedia of Experiments Archive

EDUCATION

JoVE Core JoVE Business JoVE Science Education JoVE Lab Manual Faculty Resource Center Faculty Site

Terms & Conditions of Use

Consistency Graph Modeling for Semantic Correspondence | JoVE Visualize

Home
Research Domains
Information And Computing Sciences

Consistency Graph Modeling for Semantic Correspondence

Jianfeng He, Tianzhu Zhang, Yuhui Zheng

IEEE Transactions on Image Processing : a Publication of the IEEE Signal Processing Society|May 7, 2021

View abstract on PubMed

Summary

This study introduces a new deep learning model for semantic correspondence, jointly using inter-image, intra-image, and cycle consistency information. The Consistency Graph Modeling Network (CGMNet) improves feature representation for better image matching.

Area of Science:

Computer Vision
Machine Learning
Artificial Intelligence

Background:

Establishing semantic correspondence between images of similar objects is crucial for various computer vision tasks.
Existing methods often fail to integrate different types of relational information effectively.

Purpose of the Study:

To propose a novel end-to-end deep learning model, the Consistency Graph Modeling Network (CGMNet), for robust semantic correspondence.
To jointly model inter-image relationships, intra-image relationships, and cycle consistency within a unified framework.

Main Methods:

Developed CGMNet, a deep model integrating three key modules: cross-graph, intra-graph, and cycle consistency.
The network learns discriminative feature representations robust to ambiguities and background clutter.

Main Results:

CGMNet successfully models inter-image, intra-image, and cycle consistency information simultaneously.
The model demonstrates superior performance compared to state-of-the-art methods on challenging datasets (PF-PASCAL, PF-WILLOW, Caltech-101, TSS).

Conclusions:

The proposed CGMNet is the first deep model to jointly leverage all three types of information for semantic correspondence.
This unified approach enhances feature representation, leading to improved accuracy in semantic correspondence tasks.

Data Management And Data Science

Query Processing And Optimisation

Consistency Graph Modeling For Semantic Correspondence