Logo InFlux and InFlux++

Leaderboard

Compare highlighted camera-intrinsics and LUT-Reliable EPE recall metrics. Select an evaluation set, then click the ↕, ▲, or ▼ arrow beside a metric threshold to choose the sort metric and direction. A submission not evaluated for the selected set remains visible with em dashes.

Higher recall is better ↑

InFlux — Original InFlux real-world evaluation set
Sorted by LUT-Reliable EPE Recall @300 px, highest first. 8 public results.
Method fx Recall ↑ fy Recall ↑ cx Recall ↑ cy Recall ↑ LUT-Reliable EPE Recall ↑ Publication Code Date
@1% @10% @20% @1% @10% @20% @0.5% @1% @2% @0.5% @1% @2% @10 px @50 px @300 px
AnyCalib (finetuned by InFlux++)[1] 2.21 17.5 34.1 2.23 18.7 34.9 65.5 90.0 98.5 47.4 76.1 98.3 2.09 27.9 82.7 [paper] [code] 2026-07-25
AnyCalib[2] 1.30 11.5 20.8 1.38 11.7 21.0 73.5 94.4 99.6 51.3 81.7 99.2 2.95 34.1 80.1 [paper] [code] 2026-07-25
GeoCalib[5] 5.47e-1 5.41 11.2 5.47e-1 5.40 11.2 93.4 99.3 100 84.7 91.2 100 9.32e-1 14.4 59.8 [paper] [code] 2026-07-25
WildCamera[8] 7.39e-1 7.59 17.0 6.02e-1 5.85 14.2 4.27 8.76 20.2 5.62 11.4 22.1 1.85e-1 3.64 53.7 [paper] [code] 2026-07-25
UniDepth v2[7] 3.73e-1 5.91 12.6 3.54e-1 5.79 12.1 39.6 50.4 63.2 7.82 19.7 57.0 6.20e-1 7.25 52.4 [paper] [code] 2026-07-25
DroidCalib[4] 2.71e-1 1.82 6.77 1.67e-1 1.51 4.74 4.58 7.29 16.3 3.63 6.72 15.0 3.62e-1 7.15 31.7 [paper] [code] 2026-07-25
Perspective Fields[6] 1.25e-1 1.61 4.27 1.24e-1 1.60 4.27 1.67 3.31 6.58 1.63 3.23 6.46 1.34e-2 5.49e-1 19.9 [paper] [code] 2026-07-25
COLMAP[3] 2.56e-1 1.94 3.16 2.57e-1 1.94 3.16 7.44 7.94 8.00 6.11 6.87 8.00 2.24e-1 3.59 8.89 [paper] [code] 2026-07-25

Only results explicitly published by the original submitter are shown. Submitter email addresses and private prediction files are never exposed here.

References

Reference numbers are sorted by submission time.

  1. [1] InFlux++: Real and Synthetic Data for Estimating Dynamic Camera Intrinsics. [paper] [code]
  2. [2] AnyCalib: On-Manifold Learning for Model-Agnostic Single-View Camera Calibration. [paper] [code]
  3. [3] Structure-from-Motion Revisited. [paper] [code]
  4. [4] Deep Geometry-Aware Camera Self-Calibration from Video. [paper] [code]
  5. [5] GeoCalib: Learning Single-image Calibration with Geometric Optimization. [paper] [code]
  6. [6] Perspective Fields for Single Image Camera Calibration. [paper] [code]
  7. [7] UniDepthV2: Universal Monocular Metric Depth Estimation Made Simpler. [paper] [code]
  8. [8] Tame a Wild Camera: In-the-Wild Monocular Camera Calibration. [paper] [code]