Reproduce “Cross-dimensional Weighting for Aggregated Deep Convolutional Features” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Yannis Kalantidis, Clayton Mellina, Simon Osindero
CroW creates a compact image descriptor by reweighting a pretrained CNN’s spatial feature responses according to both location and channel, then summing them. It is a non-parametric, training-free extension of pooled features. On Paris, its 256-dimensional descriptor reaches 0.765 mAP, versus 0.739 for unweighted pooling, in the setting without Oxford100k distractors or query expansion; this supports improvement over that baseline.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.