AI Paper Atlas

Research map / 3D vision / Depth and stereo

Co-Attention for Conditioned Image Matching

Olivia Wiles, Sebastien Ehrhardt, Andrew Zisserman

CVPR 2021 · 3D vision › Depth and stereo · approach: Dense feature matching for correspondence · 1 citations (OpenAlex)

Check your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Add CoAM conditioning to the CAPSNet descriptor modelCAPSNet · MegaDepth pose prediction, especially hard viewpoint pairscorrect rotational predictions (%): 65.8 -> 69.3Improved
Inject CoAM at both coarse and fine resolutions instead of coarse only (ablation)CAPSNet · MegaDepth pose predictioncorrect rotational predictions (%): easy: 91.7 -> 91.9; medium: 82.9 -> 82.8; hard: 69.3 -> 68.8Mixed
Add the distinctiveness score to matching (ablation)CD-UNet · YFCC100MmAP@5°: 31.60 -> 55.58Improved
Add CoAM conditioning (ablation)CD-UNet · YFCC100MmAP@5°: 53.43 -> 55.58Improved
Replace ResNet50 with EfficientNet-B1 (ablation)CD-UNet · YFCC100MmAP@5°: 55.58 -> 57.23Improved
Replace hinge loss with InfoNCE (ablation)CD-UNet · Aachen Day-Nightaccuracy at 0.5m (5°): 70.4 -> 62.2Hurt
Increase test-time grid size from G=128 to G=256 (ablation)CD-UNet · Aachen Day-Nightaccuracy at 0.5m (5°): 70.4 -> 68.4No clear effect
Use the local-neighborhood correspondence refinement strategy (ablation)CD-UNet · HPatchesmatching accuracyImproved
Use EfficientNet-B1 instead of ResNet50 (ablation)CD-UNet · HPatchesmatching accuracyImproved
Use pair-conditioned matching rather than SIFT features3D reconstruction pipeline · Large SfM: landmark datasetsregistered images: 500 -> 702; 1035 -> 1072; 804 -> 967Improved
Use pair-conditioned matching rather than SIFT features3D reconstruction pipeline · Large SfM: landmark datasetsreprojection error (px): 0.60 -> 1.30; 0.69 -> 1.34; 0.61 -> 1.32Hurt
Use correspondence-based stylization instead of semantic stylizationStylization model · MegaDepthmean L1 error: 0.22 -> 0.14Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy