Issues

• XFeat performs significantly worse than SuperPoint on repetitive textures

  verlab/accelerated_features#121 · Filter by author DoMoCo opened on May 14, 2026
  1

• Error in xfeat+lg_torch_hub.ipynb

  verlab/accelerated_features#119 · Filter by author lcmrl opened on Feb 18, 2026
  1

• Batch inference not faster than single inference

  verlab/accelerated_features#117 · Filter by author octaAIteam opened on Oct 1, 2025

• XFeat + Lightglue Training

  verlab/accelerated_features#116 · Filter by author 1303463822 opened on Sep 29, 2025
  1

• Synthetic-only training Issue

  verlab/accelerated_features#115 · Filter by author htuann2712 opened on Sep 18, 2025

• About InterpolateSparse2d in interpolator.py

  verlab/accelerated_features#114 · Filter by author IkariKlein opened on Aug 19, 2025

• Clarification on MegaDepth Dataset Preparation & Broken Link

  verlab/accelerated_features#113 · Filter by author skbhat opened on Aug 14, 2025
  2

• Question about the smaller model

  verlab/accelerated_features#112 · Filter by author jirongli opened on Aug 6, 2025

• request for experiment from glue-factory

  verlab/accelerated_features#111 · Filter by author Vas-sA opened on Jul 30, 2025

• If you are interested in image matching, I have created a WeChat group. Please add my personal WeChat account, 13064718889, and indicate image matching

  verlab/accelerated_features#110 · Filter by author canquanshen opened on Jul 17, 2025

• about batchsize>1, the number of keypoints

  verlab/accelerated_features#108 · Filter by author July0928 opened on Jul 8, 2025

• when the small image size match big image size ,the result of match is bad

  verlab/accelerated_features#107 · Filter by author canquanshen opened on Jul 2, 2025

• an error in xfeat+lg_torch_hub.ipynb

  verlab/accelerated_features#106 · Filter by author HumanDr opened on Jun 23, 2025

• TPS transformations at the time of data enhancement

  verlab/accelerated_features#105 · Filter by author Smile-lll opened on May 27, 2025

• Question about input first two layers, mean + InstanceNorm2d

  verlab/accelerated_features#104 · Filter by author noahzhy opened on May 12, 2025

• Megadepth dataset and training convergence

  verlab/accelerated_features#103 · Filter by author Odin-byte opened on May 6, 2025
  4

• how to convert lightglue to RKNN format, thank you

  verlab/accelerated_features#102 · Filter by author anheidadi opened on Apr 2, 2025
  3

• The network is not convergence when training.

  verlab/accelerated_features#101 · Filter by author tangconghui2008 opened on Mar 21, 2025
  3

• [Feature Inquiry] ONNX export support for lighterglue model

  verlab/accelerated_features#100 · Filter by author Teddywesside1 opened on Mar 12, 2025
  12

• The problem of convergence time in training the xfeat model

  verlab/accelerated_features#99 · Filter by author codekiller000 opened on Feb 26, 2025

• How to Use XFeat in a Web Environment (WebGL/WASM)?

  verlab/accelerated_features#96 · Filter by author NithinJangid opened on Feb 19, 2025

• Problem when training xfeat+lightglue

  verlab/accelerated_features#94 · Filter by author limymy opened on Jan 26, 2025
  5

• I find a bug when trying xfeat with lightglue

  verlab/accelerated_features#93 · Filter by author ZhouShunlong opened on Jan 10, 2025

• How do I make my own datasets to train？

  verlab/accelerated_features#90 · Filter by author dengxiongshi opened on Dec 24, 2024

• The issue of obtaining the weight file for xfeat-lighterglue.pt.

  verlab/accelerated_features#89 · Filter by author 364584602 opened on Dec 23, 2024