What AI says about Neural Relation Graph

The repository from SNU ML Lab provides the official PyTorch implementation of Neural Relation Graph, a unified framework for identifying label noise and outlier data introduced at NeurIPS 2023. It includes code, model files, and setup guidance (dependencies like PyTorch 1.11 and timm 0.5.4), plus instructions for running data processing and experiments (e.g., ImageNet data and Google Drive checkpoints). The project is open-source under the MIT license and serves as a reference implementation for detecting mislabeled data and outliers using neural relation graphs.