Bug report (I searched for similar issues and did not find one
Current behavior
I used WinML Dashboard to overwrite Image.NomimalPixelRange to range [-1.1]. But it has no effect in WinML wiml::NominalRangeConverter::NominalRangeConverter (still defaulting to 0-255). No meta data values if loaded in Netro too.
Expected behavior
Metadata values shall be correctly injected.
Minimal reproduction of the problem with instructions
Take a ONNX model. Edit and overwrite Image.NomimalPixelRange to range [-1.1. Then, load into a sample (for example, BatchSupport), Add a breakpoint in wiml::NominalRangeConverter::NominalRangeConverter and check which value is loaded in pixelRange.
Environment
Windows Build Number: 19043.1237
App min and target version:
commit: 9a39f9ea (0.7 pre-release)
OS Version (Server, IoT Core, Desktop, etc):
Graphics Driver version:
DxDiag:
WinMLTools specific:
Source training framework: (e.g. CoreML, Scikit-learn, …)
WinMLTools version : WinML Dashboard : 0.7 Pre-release [9a39f9ea ]
I'm submitting a…
Bug report (I searched for similar issues and did not find one
Current behavior
I used WinML Dashboard to overwrite Image.NomimalPixelRange to range [-1.1]. But it has no effect in WinML wiml::NominalRangeConverter::NominalRangeConverter (still defaulting to 0-255). No meta data values if loaded in Netro too.
Expected behavior
Metadata values shall be correctly injected.
Minimal reproduction of the problem with instructions
Take a ONNX model. Edit and overwrite Image.NomimalPixelRange to range [-1.1. Then, load into a sample (for example, BatchSupport), Add a breakpoint in wiml::NominalRangeConverter::NominalRangeConverter and check which value is loaded in pixelRange.
Environment
Windows Build Number: 19043.1237
App min and target version: commit: 9a39f9ea (0.7 pre-release) OS Version (Server, IoT Core, Desktop, etc):
Graphics Driver version:
DxDiag:
WinMLTools specific:
Visual Studio