AI Paper Atlas

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

Interpretable Image Classification with Differentiable Prototypes Assignment

Dawid Rymarczyk, Łukasz Struski, Michał Górszczak, Koryna Lewandowska, Jacek Tabor, Bartosz Zieliński

ECCV 2022 · Label-efficient and robust learning › Recognition and long-tail · approach: Self-explaining and visual explanation networks · 119 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
add orthogonalization loss (ablation)ProtoPool · CUB-200-2011accuracy: 82.4 -> 85.5Improved
add orthogonalization loss (ablation)ProtoPool · Stanford Carsaccuracy: 86.8 -> 88.9Improved
use Gumbel-Softmax instead of softmax (ablation)ProtoPool · CUB-200-2011accuracy: 80.3 -> 85.5Improved
use Gumbel-Softmax instead of softmax (ablation)ProtoPool · Stanford Carsaccuracy: 64.5 -> 88.9Improved
add Gumbel-Softmax and orthogonalization loss together (ablation)ProtoPool · CUB-200-2011accuracy: 65.1 -> 85.5Improved
add Gumbel-Softmax and orthogonalization loss together (ablation)ProtoPool · Stanford Carsaccuracy: 30.8 -> 88.9Improved
use focal similarity instead of ProtoPNet similarity (ablation)ProtoPool · CUB-200-2011accuracy: 85.3 -> 85.5No clear effect
use focal similarity instead of ProtoPNet similarity (ablation)ProtoPool · Stanford Carsaccuracy: 88.8 -> 88.9No clear effect
apply prototype projection (ablation)ProtoPool ResNet34 · CUB-200-2011accuracy: 80.8 -> 80.3No clear effect
apply prototype projection (ablation)ProtoPool ResNet50 · Stanford Carsaccuracy: 88.4 -> 88.9No clear effect
use ProtoPool instead of ProtoPShare with fewer prototypesResNet34 · CUB-200-2011accuracy: 74.7 -> 80.3Improved
use ProtoPool instead of TesNet with 10x fewer prototypesResNet34 · Stanford Carsaccuracy: 92.6 -> 89.3Hurt

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