Reproduce “Robust Self-Paced Hashing for Cross-Modal Retrieval with Noisy Labels” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Ruitao Pu, Yuan Sun, Yang Qin, Zhenwen Ren, Xiaomin Song, Huiming Zheng, Dezhong Peng
Researchers from Sichuan University introduced Robust Self-paced Hashing with Noisy Labels (RSHNL), an AI-driven method that significantly improves cross-modal retrieval by identifying and mitigating the impact of noisy labels. It consistently outperforms 11 state-of-the-art baselines across various noise rates and hash bit lengths on multiple datasets, demonstrating enhanced robustness.
Key insights
- Explicitly distinguishing and assigning zero weight to noisy labels, guided by a self-paced learning mechanism, is highly effective for robust cross-modal hashing.
- A pre-training or "warm-up" phase helps stabilize model parameters, making the subsequent noise-tolerance learning curriculum more effective in identifying and handling corrupted labels.
- The progressive "easy-to-hard" learning curriculum, combined with inter-modal consistency (CHL) and intra-class compactness (CAL), allows for the generation of more discriminative and generalizable hash codes even with significant label noise.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.