Visual loop closure detection is an important module in visual simultaneous localization and mapping (SLAM), which associates current camera observation with previously visited places. Loop closures correct drifts in trajectory estimation to build a globally consistent map. However, a false loop closure can be fatal, so verification is required as an additional step to ensure robustness by rejecting the false positive loops. Geometric verification has been a well-acknowledged solution that leverages spatial clues provided by local feature matching to find true positives. Existing feature matching methods focus on homography and pose estimation in long-term visual localization, lacking references for geometric verification. To fill the gap, this paper proposes a unified benchmark targeting geometric verification of loop closure detection under long-term conditional variations. Furthermore, we evaluate six representative local feature matching methods (handcrafted and learning-based) under the benchmark, with in-depth analysis for limitations and future directions.

• ⭐ Paper is release on arxiv.
• 🎉 The paper is accepted by IROS 2024!
• 🚀 Releasing the visualization of image matching results. (google drive)
• 🚀 Releasing the benchmark (easy)! Checkout the image pairs from dataset/gt and images from google drive

We use part of the HLoc code for feature extraction and matching.

git clone && cd GV-Bench
git submodule init
git submodule update
cd third_party/Hierarchival-Localization
git checkout gvbench # this is a customized fork version
python -m pip install -e .

We provide the output results with the format shown below. You can use these results directly.

$seq_$feature_$match.log
$seq_$feature_$match.npy # with following format

np.save(str(export_dir), {
  'prob': num_matches_norm,
  'qImages': qImages,
  'rImages': rImages,
  'gt': labels, 
  'inliers': inliers_list,
  'all_matches': pointMaps,
  'precision': precision, 
  'recall': recall, 
  'TH': TH,
  'average_precision': average_precision,
  'Max Recall': r_recall
  })

To get standard feature detection and matching results, we proposed to use hloc.

• Download the dataset sequences from google drive and put it under the dataset/ folder.

• Extract and match feature using hloc.

  • Extract features: SIFT, SuperPoint, and DISK

    python third_party/Hierarchical-Localization/gvbench_utils.py config/${seq}.yaml --extraction 

  • Match features: SIFT-NN, SIFT-LightGlue (Not yet implemented), SuperPoint-NN, DISK-NN, SuperPoint-SuperGlue, SuperPoint-LightGlue, DISK-LightGlue, LoFTR

    # all methods except LoFTR
    python third_party/Hierarchical-Localization/gvbench_utils.py config/${seq}.yaml --matching
    
    # LoFTR is different from above methods thus
    python third_party/Hierarchical-Localization/gvbench_utils.py config/${seq}.yaml --matching_loftr

  • Image pairs files
    • We prepare pairs (GT) file for matching under dataset/gt foler.
    • Make sure to use the fork hloc for feature extraction and matching https://github.com/jarvisyjw/Hierarchical-Localization.git -b gvbench

• Evaluation

  • We provide out-of-box scripts

  cd GV-Bench/scripts
  bash ./evaluation <day> # run script with 
  #sequence name: day, night, night-hard, season, season-hard, weather

• Using customized local features for geometric verification (GV).

  TODO

• Visualization
  • Demos are presented in plot_data.ipynb

• This work builds upon hloc, thanks for their amazing work.
• Contact: jingwen.yu@connect.ust.hk