YOLO9000: Better, Faster, Stronger(一)论文地址:(二)核心思想:(三)YOLO 的不足:(四)Batch Normalization:(五)High Resolution Classifier:(六)Convolutional With Anchor Boxes:(七)Dimension Clusters(八)Fine-Grained Features:(九)Multi-Scale Training:(十)Darknet-19:(十一)Joint Training:(十二)实验结果: (一)论文地址: https://arxiv.org/abs/1612.0