On COCO validation with matched baseline Faster R-CNN, replacing VGG-16 by ResNet-101 raises mAP@0.5 from 41.5% to 48.4% and COCO-standard mAP@[.5,.95] from 21.2% to 27.2%. The latter is reported as a 6.0-point absolute and 28% relative improvement, nearly as large in absolute terms as the 6.9-point mAP@0.5 gain.

On COCO validation with matched baseline Faster R-CNN, replacing VGG-16 by ResNet-101 raises mAP@0.5 from 41.5% to 48.4% and COCO-standard mAP@[.5,.95] from 21.2% to 27.2%. The latter is reported as a 6.0-point absolute and 28% relative improvement, nearly as large in absolute terms as the 6.9-point mAP@0.5 gain. · CiteArk

测量结果

已评估 0/7 项

获得支持 0有争议或冲突 0尚未评估 7

保留同一结论下的完整结果组。已有评估排在前面；单项判断不替代整组结论。展开行查看条件、来源与历次证据。

每行是一项测量，分别显示论文报告、最近可用实测和累计证据状态。
论文报告	最近可用实测	证据判断
41.5%	—	尚未评估
21.2%	—	尚未评估
48.4%	—	尚未评估
27.2%	—	尚未评估
6.9 percentage points	—	尚未评估
6 percentage points	—	尚未评估
28%	—	尚未评估

报告指标与观测值

Assessment（0）

关联运行（0）

测量结果