map0.5和map0.95

mAP@0.5表示在IoU阈值为0.5时的平均准确率。这意味着只有当检测框与真实目标的重叠区域大于等于50%时，才会将其视为正确的检测结果。mAP@0.95表示在IoU阈值为0.95时的平均准确率。这意味着只有当检测框与真实目标的重叠区域大于等于95%时，才会将其视为正确的检测结果。根据引用和，mAP@0.95的数值通常会比mAP@0.5的数值要小，因为较高的IoU阈值要求更严格的检测准确性。在Yolo系列中，mAP@0.5的效果通常较好，但mAP@0.95的效果较差。可能是因为过高的IoU阈值对于检测框的位置要求过高，同时，ground truth（真实标注）也可能存在一些误差，导致无法达到高精度性。<span class="em">1</span><span class="em">2</span><span class="em">3</span>
#### 引用[.reference_title]
- *1* [mAP@0.5 与 mAP@0.5:0.95 是什么意思，YOLOv5](https://blog.csdn.net/qq_38251616/article/details/122109519)[target="_blank" data-report-click={"spm":"1018.2226.3001.9630","extra":{"utm_source":"vip_chatgpt_common_search_pc_result","utm_medium":"distribute.pc_search_result.none-task-cask-2~all~insert_cask~default-1-null.142^v92^chatsearchT0_1"}}] [.reference_item style="max-width: 33.333333333333336%"]
- *2* [【深度学习】目标检测的性能评价指标，mAP_0.5，mAP_0.5,0.95,0.05](https://blog.csdn.net/x1131230123/article/details/127740011)[target="_blank" data-report-click={"spm":"1018.2226.3001.9630","extra":{"utm_source":"vip_chatgpt_common_search_pc_result","utm_medium":"distribute.pc_search_result.none-task-cask-2~all~insert_cask~default-1-null.142^v92^chatsearchT0_1"}}] [.reference_item style="max-width: 33.333333333333336%"]
- *3* [mAP@0.5与mAP@0.5:0.95的含义](https://blog.csdn.net/frighting_ing/article/details/121197733)[target="_blank" data-report-click={"spm":"1018.2226.3001.9630","extra":{"utm_source":"vip_chatgpt_common_search_pc_result","utm_medium":"distribute.pc_search_result.none-task-cask-2~all~insert_cask~default-1-null.142^v92^chatsearchT0_1"}}] [.reference_item style="max-width: 33.333333333333336%"]
[ .reference_list ]