AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

Convolutional Patch Representations for Image Retrieval: An Unsupervised Approach

Mattis Paulin, Julien Mairal, Matthijs Douze, Zaïd Harchaoui, Florent Perronnin, Cordelia Schmid

IJCV 2016 · Label-efficient and robust learning › Recognition and long-tail · approach: Correspondence and causal representation learning · 64 citations (OpenAlex)

Read the paperCheck 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
retrain PhilippNet on 100K real Rome classes instead of surrogate classes (ablation)PhilippNet · RomePatches-train patch retrievalmAP: 86.1 -> 89.9Improved
retrain PhilippNet on only 10K Rome classes (ablation)PhilippNet · RomePatches-train patch retrievalmAP: 86.1 -> 84.1Hurt
use CKN-grad unsupervised instead of supervised CNNsPhilippNet-Rome 100K · RomePatches-test patch retrievalmAP: 85.6 -> 88.1Improved
fine-tune AlexNet on Landmarks instead of ImageNet (conv2) (ablation)AlexNet · Oxford image retrievalmAP: 12.5 -> 23.8Improved
fine-tune AlexNet on Landmarks instead of ImageNet (conv3) (ablation)AlexNet · Holidays image retrievalmAP: 79.3 -> 76.1Hurt
fine-tune AlexNet on Landmarks instead of ImageNet (conv3) (ablation)AlexNet · Oxford image retrievalmAP: 33.3 -> 38.6Improved
concatenate VLADs of the three CKN channels (mix) versus CKN-white (ablation)CKN · Holidays image retrievalmAP: 78.7 -> 79.3Improved
PCA+whitening to 4096 dimsCKN-mix · Holidays image retrievalmAP: 79.3 -> 82.9Improved
PCA+whitening to 4096 dimsCKN-grad · Oxford image retrievalmAP: 49.8 -> 47.2Hurt
use dense keypoints with increased pooling instead of Hessian-AffineCKN-white · Holidays image retrievalmAP: 78.7 -> 80.8Improved
use dense keypoints with unchanged pooling instead of Hessian-Affine (ablation)CKN-white · Holidays image retrievalmAP: 78.7 -> 76.8Hurt
retune CKN-grad hyperparameters (fewer bins, more pooling) for DoG patchesCKN-grad · Liberty pairsFPR@95% recall: 27.7 -> 14.4Improved

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