Learning to Produce Semi-dense Correspondences for Visual Localization

1 code implementation • 13 Feb 2024 • Khang Truong Giang, Soohwan Song, Sungho Jo

To tackle this issue, we propose a novel localization method that extracts reliable semi-dense 2D-3D matching points based on dense keypoint matches.

Pose Estimation Visual Localization

TopicFM+: Boosting Accuracy and Efficiency of Topic-Assisted Feature Matching

1 code implementation • 2 Jul 2023 • Khang Truong Giang, Soohwan Song, Sungho Jo

By incorporating these topics, we can effectively capture comprehensive context information and obtain discriminative and high-quality features.

Computational Efficiency

TopicFM: Robust and Interpretable Topic-Assisted Feature Matching

1 code implementation • 1 Jul 2022 • Khang Truong Giang, Soohwan Song, Sungho Jo

This study addresses an image-matching problem in challenging cases, such as large scene variations or textureless scenes.

Curvature-guided dynamic scale networks for Multi-view Stereo

1 code implementation • ICLR 2022 • Khang Truong Giang, Soohwan Song, Sungho Jo

As a result, the method can process higher resolution inputs within faster run-time and lower memory than other MVS methods.

3D Reconstruction Depth Estimation +1