Merge pull request #21025 from alalek:issue_21004
* dnn(ocl4dnn): fix LRN layer accuracy problems - FP16 intermediate computation is not accurate and may provide NaN values * dnn(test): update tolerance for FP16
Showing
想要评论请 注册 或 登录
* dnn(ocl4dnn): fix LRN layer accuracy problems - FP16 intermediate computation is not accurate and may provide NaN values * dnn(test): update tolerance for FP16