diff options
| author | James Conroy <james.conroy@arm.com> | 2020-02-20 11:14:49 +0000 |
|---|---|---|
| committer | James Conroy <james.conroy@arm.com> | 2020-02-20 15:08:21 +0000 |
| commit | ed966716889cfd614d310432d84804cc77aef108 (patch) | |
| tree | d1896be26f73268098f6d126f3e797cbd60cbd01 /src/backends/neon/workloads/NeonDetectionPostProcessWorkload.cpp | |
| parent | a90573581bb2f63f229ddd61650163b9fc0fdd3e (diff) | |
| download | armnn-ed966716889cfd614d310432d84804cc77aef108.tar.gz | |
IVGCVSW-4456 Fix MobileNet SSD perf regression
* Now use accelerated DetectionPostProcess
NEON workload, originally was using vanilla
version.
Change-Id: Ib64653801024a191f841d601932fa3c54baa5c3b
Signed-off-by: James Conroy <james.conroy@arm.com>
Diffstat (limited to 'src/backends/neon/workloads/NeonDetectionPostProcessWorkload.cpp')
| -rw-r--r-- | src/backends/neon/workloads/NeonDetectionPostProcessWorkload.cpp | 2 |
1 files changed, 1 insertions, 1 deletions
diff --git a/src/backends/neon/workloads/NeonDetectionPostProcessWorkload.cpp b/src/backends/neon/workloads/NeonDetectionPostProcessWorkload.cpp index 481e950953..2ed47e4463 100644 --- a/src/backends/neon/workloads/NeonDetectionPostProcessWorkload.cpp +++ b/src/backends/neon/workloads/NeonDetectionPostProcessWorkload.cpp @@ -62,7 +62,7 @@ arm_compute::Status NeonDetectionPostProcessValidate(const TensorInfo& boxEncodi arm_compute::TensorInfo aclNumDetections = armcomputetensorutils::BuildArmComputeTensorInfo(numDetections); - return arm_compute::CPPDetectionPostProcessLayer::validate( + return arm_compute::NEDetectionPostProcessLayer::validate( &aclBoxEncodings, &aclScores, &aclAnchors, |