Dynamic deep multi-label image data augmentation based on self-paced learning

Bin Yu

doi:10.1016/j.cviu.2025.104530

Dynamic deep multi-label image data augmentation based on self-paced learning

2025-10-14

0

PRE

AI

DOI：10.1016/j.cviu.2025.104530

Original

Original request for help

Save

Abstract

En 中文

• End-to-end framework improves multi-label classification via adaptive augmentation. • Self-paced learning selects easy samples to guide augmentation during training. • Image data synthesis reduces class imbalance in self-paced selected samples. • Noise-mitigation filters low-similarity samples to improve training quality.

Dynamic deep multi-label image data augmentation based on self-paced learning

Abstract

Journal

Organization