AI Paper Atlas

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

Weakly Supervised Fine-Grained Image Classification via Guassian Mixture Model Oriented Discriminative Learning

Zhihui Wang, Shijie Wang, Shuhui Yang, Haojie Li, Jianjun Li, Zezhou Li

CVPR 2020 · Label-efficient and robust learning › Recognition and long-tail · approach: Correspondence and causal representation learning · 101 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 score-based default-patch sampling to the ResNet-50 baseline. (ablation)BL · CUB-200-2011accuracy: 84.5% -> 86.2%Improved
Add DF-GMM to score-based patch sampling. (ablation)BL + Sample · CUB-200-2011accuracy: 86.2% -> 88.8%Improved
Add score-based sampling to the baseline. (ablation)BL · CUB-200-2011speed: n/a -> 50 fpsImproved
Add DF-GMM to score-based patch sampling. (ablation)BL + Sample · CUB-200-2011speed: 50 fps -> 41 fpsHurt
Use global average pooling instead of random initialization for low-rank bases. (ablation)DF-GMM · CUB-200-2011accuracy: 87.1% -> 88.8%Improved
Use global average pooling instead of global max pooling for low-rank base initialization. (ablation)DF-GMM · CUB-200-2011accuracy: 87.9% -> 88.8%Improved
Increase the GMM iteration count from 1 to 3. (ablation)DF-GMM · CUB-200-2011accuracy: 86.9% -> 88.8%Improved
Increase the GMM iteration count from 2 to 3. (ablation)DF-GMM · CUB-200-2011accuracy: 87.5% -> 88.8%Improved
Increase the GMM iteration count from 3 to 4. (ablation)DF-GMM · CUB-200-2011accuracy: 88.8% -> 88.4%Hurt
Increase the GMM iteration count from 3 to 5. (ablation)DF-GMM · CUB-200-2011accuracy: 88.8% -> 88.1%Hurt
Increase the number of selected discriminative patches from 2 to 4. (ablation)DF-GMM · CUB-200-2011accuracy: 88.10% -> 88.80%Improved
Increase the number of selected discriminative patches from 2 to 4. (ablation)DF-GMM · CUB-200-2011speed: 43.00 -> 41.00Hurt

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