Home

Awesome

Noisy-Correspondence Learning Summary (Updating)

A new research direction of label noise learning. Noisy correspondence learning was first formally proposed in NCR (NeurIPS 2021, Oral) by the XLearning group, which aims to eliminate the negative impact of the mismatched pairs (e.g., false positives/negatives) instead of annotation errors in several tasks.

We mark works contributed by ourselves with ⭐.

This repository now is maintained by Mouxing Yang, Yijie Lin, Changhao He, and Yang Qin. We hope more AI-workers join us and thank all contributors!

Tasks

Image-Text Matching/RetrievalVision-Language Pre-training
Re-identificationVideo-Text Learning
Image CaptioningImage Contrastive Learning
Graph Matching Visual-Audio Learning
Machine Reading ComprehensionDense Retrieval
Multi-View Clustering

Image-Text Matching/Retrieval

2024

2023

2022

2021

Vision-Language Pre-training

Re-identification

Video-Text Learning

Image Captioning

Image Contrastive Learning

Graph Matching

Visual-Audio Learning

Machine Reading Comprehension

Dense Retrieval

Multi-View Clustering

2024

2023

2022

2021

2020