diff options
author | Michele Di Giorgio <michele.digiorgio@arm.com> | 2018-11-16 10:02:26 +0000 |
---|---|---|
committer | Michele Di Giorgio <michele.digiorgio@arm.com> | 2018-11-16 17:47:02 +0000 |
commit | c8df89f477c3dc63f396ad37bee8ed5d50dee4ac (patch) | |
tree | 4fc4151b382438416b531b3bcb586f88eb32c2d1 /src/graph | |
parent | a25d16c86f0d870408bc8b941aa755093417b0f0 (diff) | |
download | ComputeLibrary-c8df89f477c3dc63f396ad37bee8ed5d50dee4ac.tar.gz |
COMPMID-1451: (3RDPARTY_UPDATE) Fixes for GenerateProposals graph node and BoxWithNMSLimitKernel
COMPMID-1792: Accuracy issue in CLGenerateProposals
This patch does the following:
- Some fixes for GenerateProposals function and tests
- Adapting BoxWithNMSLimitKernel to only accept U32 tensors as keeps_size
- Update 3rdparty
- Adds a small tolerance for a GenerateProposals test
Change-Id: Ia8ec1cdfe941fe05003645e86deb9ea6a6044d74
Diffstat (limited to 'src/graph')
-rw-r--r-- | src/graph/nodes/GenerateProposalsLayerNode.cpp | 3 |
1 files changed, 2 insertions, 1 deletions
diff --git a/src/graph/nodes/GenerateProposalsLayerNode.cpp b/src/graph/nodes/GenerateProposalsLayerNode.cpp index f5a3c02dd5..7367e80539 100644 --- a/src/graph/nodes/GenerateProposalsLayerNode.cpp +++ b/src/graph/nodes/GenerateProposalsLayerNode.cpp @@ -80,7 +80,8 @@ TensorDescriptor GenerateProposalsLayerNode::configure_output(size_t idx) const break; case 2: // Configure num_valid_proposals - output_desc.shape = TensorShape(1); + output_desc.shape = TensorShape(1); + output_desc.data_type = DataType::U32; break; default: ARM_COMPUTE_ERROR("Unsupported output index"); |