AI Paper Atlas

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

Examining CNN Representations With Respect to Dataset Bias

Quanshi Zhang, Wenguan Wang, Song-Chun Zhu

AAAI 2018 · Label-efficient and robust learning › Recognition and long-tail · approach: Data augmentation and generalization bias · 96 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
Use mined representation relationships rather than annotation entropy to select top-10 failure modesAlexNet · CelebAdecrease of accuracy: 10.37 vs 25.85Improved
Use mined representation relationships rather than annotation entropy to select top-10 failure modesAlexNet · CelebAaccuracy in ordinary images: 69.22 vs 72.29Mixed
Use mined representation relationships rather than annotation entropy to select top-40 failure modesAlexNet · SUN Attributedecrease of accuracy: 27.47 vs 29.68Improved
Use mined representation relationships rather than annotation entropy to select top-40 failure modesAlexNet · SUN Attributeaccuracy in ordinary images: 63.36 vs 68.65Mixed
Reduce training-set annotation bias by adding samples with opposite attribute annotationsAlexNet trained on manually biased datasets · CelebA and SUN AttributeKL divergence of attribute relationshipsImproved
Use estimated representation relationships instead of annotation co-occurrence alone to identify failure modesAlexNet · CelebA and SUN Attributeaccuracy decrease on identified failure modesImproved

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