研究人员推出了一种新的网络监督多标签识别基准，该领域使用免费的网络图像来训练深度学习模型，从而减少了对昂贵手动标注的需求。该基准名为 Web-COCO 和 Web-Pascal，包含约 300,000 张图像，旨在标准化多标签识别任务的评估协议。除了基准之外，研究团队还提出了一个双分支多标签对比学习 (DBMLCL) 框架，该框架在识别和纠正噪声标签方面表现出卓越的性能。 AI
影响 这项工作可以通过利用现成可用的网络数据，从而更高效、更经济地训练图像识别模型。
排序理由 该集群描述了一篇介绍特定机器学习任务的基准和新颖框架的学术论文。
在 Hugging Face Daily Papers 阅读 →
AI 生成摘要 · Google Gemini · 来自 2 个来源。 我们如何撰写摘要 →
完整方法见我们的编辑标准。
Training deep learning models with freely available web images can reduce their dependence on costly manual annotations. Although webly supervised learning has been widely studied for single-label recognition, its multi-label counterpart remains underexplored, partly due to the l…
arXiv:2607.20874v1 Announce Type: new Abstract: Training deep learning models with freely available web images can reduce their dependence on costly manual annotations. Although webly supervised learning has been widely studied for single-label recognition, its multi-label counte…