AI Paper Atlas

Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning

Coarse-to-Fine Incremental Few-Shot Learning

Xiang Xiang, Yuwen Tan, Qian Wan, Jing Ma, Alan Yuille, Gregory D. Hager

ECCV 2022 · Label-efficient and robust learning › Zero-/few-shot and meta-learning · approach: Fine-grained and cross-domain few-shot baselines · 17 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 contrastive MoCo base learning (ablation)Knowe · BREEDS living17avg total accuracy: 40.10 -> 54.62Improved
add classifier-weight normalization (ablation)Knowe · BREEDS living17avg total accuracy: 30.77 -> 54.62Improved
add classifier-weight normalization (ablation)Knowe · BREEDS living17forgetting rate: 0.57 -> 0.33Improved
add freezing of memorized classifier weights (ablation)Knowe · BREEDS living17avg total accuracy: 22.09 -> 54.62Improved
add freezing of memorized classifier weights (ablation)Knowe · BREEDS living17forgetting rate: 0.66 -> 0.33Improved
add decoupling (freeze embedding weights) (ablation)Knowe · BREEDS living17avg total accuracy: 50.66 -> 54.62No clear effect
add contrastive learning to fine-tuned coarse model without weight ops (ablation)ResNet12 FT · CIFAR-100 (fine-tuning w/o IL)now accuracy: ~25% -> ~35%Improved
use Knowe instead of ANCOR fine-tuningResNet-50 · BREEDS living17avg total accuracy: 21.75 -> 54.62Improved
use Knowe instead of Weight Align+ResNet-50 · BREEDS entity30avg total accuracy: 41.78 -> 47.79Improved
use Knowe instead of Subsp. Reg.+ResNet12 · CIFAR-100 (C2FSCIL)avg total accuracy: 33.00 -> 38.50Improved
use Knowe instead of ScaILResNet12 · tieredImageNet (C2FSCIL)avg total accuracy: 32.93 -> 33.24Improved
use Knowe incremental learning instead of joint training upper boundResNet-50 · BREEDS living17avg total accuracy: 54.62 vs 55.29No clear effect

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