中文重新思考工业检测中的迁移学习：DINOv3与ImageNet预训练在RGB和X射线任务上的对比
ENRethinking Transfer Learning for Industrial Inspection: DINOv3 vs. ImageNet Pretraining Across RGB and X-ray Tasks
近期研究对比了ConvNeXt骨干网络在监督ImageNet预训练与自监督（MAE、DINOv2等）预训练对工业视觉检测的影响。结果表明，自监督预训练整体优于监督预训练，但优势不如自然图像任务显著，且最佳方法因数据集而异。这提示自监督特征可迁移，但仍需针对工业细粒度任务优化。
arXiv:2605.23472v1 Announce Type: new Abstract: Vision foundation models pretrained on web-scale data have recently shown strong transfer capabilities on many downstream tasks, but their effectiveness for industrial visual inspection remains unclear. Industrial data differ substantially from web-data and often require fine-grained dense prediction, raising the question of whether modern self-supervised pretraining can improve over the conventional transfer-learning paradigm based on supervised ImageNet initialization. In this work, we compare ConvNeXt backbones pretrained with supervised Image