kaggle-imc25 n-st Place Solution

像站在大门前倏忽了一瞬间

https://www.kaggle.com/competitions/image-matching-challenge-2025/discussion/583058

自己的知识储备和工程能力远远不够，导致无法驾驭现有的工具和方法。

作者承诺会公开代码，等待学习…

---

值得注意的点

Rotation correction

MASt3R的特征提取, 不需要像ALIKED等特征提取器一样，对图像进行旋转增强/过滤

keypoints being concentrated in only one part of the image

关键点集中于图像某一部分

https://www.kaggle.com/competitions/image-matching-challenge-2025/discussion/583401 和另外一篇discussion中都有提到这个问题。

3rd solution提出的tiled images策略很有意思。

4rd RDD

RDD: Robust Feature Detector and Descriptor using Deformable Transformer

有时间读一下这个

https://xtcpete.github.io/rdd/

https://www.kaggle.com/competitions/image-matching-challenge-2025/discussion/582959

pair-match

https://arxiv.org/abs/2504.20040

In addition to the semi-dense matches, keypoints extracted by other keypoint detectors are also fed to the MASt3R matcher (Inspired by MP-SfM https://arxiv.org/abs/2504.20040). These additional keypoints might regionally overlap with points subsampled by MASt3R, but this approach improved the score compared to using only MASt3R matches.

MASt3R subsample + ALIKED detector + SuperPoint detector