AI Paper Atlas

Research map / Label-efficient and robust learning / OOD and anomaly detection

Fast Trainable Projection for Robust Fine-tuning

Junjiao Tian, Yen-Cheng Liu, James Smith, Zsolt Kira

NeurIPS 2023 · Label-efficient and robust learning › OOD and anomaly detection · approach: Robust fine-tuning for OOD generalization · 1 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
replace TPGM with FTP (single-pass trainable projection)Projected fine-tuning · DomainNet MoCo-V3 ResNet50 100%training time (s/it): 0.80 -> 0.51Improved
replace TPGM with FTPProjected fine-tuning · DomainNet MoCo-V3 ResNet50 100%OOD avg accuracy: 36.86 -> 37.62Improved
replace TPGM with FTPProjected fine-tuning · DomainNet CLIP ResNet50 100%OOD avg accuracy: 39.65 -> 39.94Improved
replace TPGM with FTPProjected fine-tuning · DomainNet CLIP ResNet50 10%OOD avg accuracy: 30.42 -> 29.83No clear effect
replace TPGM with FTPProjected fine-tuning · DomainNet CLIP ResNet50 10%training time (s/it): 1.10 -> 0.69Improved
apply FTP instead of vanilla fine-tuningFine-tuning · DomainNet CLIP ResNet50 100%ID accuracy (Real): 80.93 -> 84.22Improved
apply FTP instead of vanilla fine-tuningFine-tuning · PASCAL semantic segmentationOOD avg mIoU: 44.00 -> 58.45Improved
apply FTP instead of vanilla fine-tuningFine-tuning · PASCAL surface normalOOD avg RMSE: 23.48 -> 20.48Improved
apply FTP instead of TPGMProjected fine-tuning · PASCAL human parts segmentationOOD avg mIoU: 51.04 -> 52.12Improved
replace TPGM with FTPProjected fine-tuning · PASCAL semantic segmentationtraining time (s/it): 0.611 -> 0.401Improved
compare MARS-SP with vanilla fine-tuningFine-tuning · PASCAL semantic segmentationOOD avg mIoU: 44.00 -> 43.59No clear effect
use FTP and linear-probe style baseline LP instead of vanilla fine-tuningFine-tuning · DomainNet CLIP ResNet50 100%ID accuracy (Real): 80.93 -> 52.56Hurt

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