Input Shape Mismatch

[gktval]

Describe the issue

I have an image classification model that was created in python. I am trying to load the onnx model in c#, but am running into errors. The image data the model was created on was 4 channel 256x256 images. When loading the model, the shape is correct, but the data length seems to be only 3 channel rgb instead of rgba (i.e. 256x256x3 = 196608). Here is the error:

System.InvalidOperationException: 'Input shape mismatch: Input 'sequential_input' has shape 1,256,256,4, but input data is of length 196608.'

Here is the code for loading the model:

var pipeline = mlContext.Transforms
.ResizeImages("image", 256, 256, inputColumnName: "sequential_input")
.Append(mlContext.Transforms.ExtractPixels("sequential_input", "image"))
.Append(mlContext.Transforms.ApplyOnnxModel("dense_1", "sequential_input", ONNX_MODEL_PATH, gpuDeviceId: 0));

var emptyDv = mlContext.Data.LoadFromEnumerable(new OnnxInput[] { });

var model = pipeline.Fit(emptyDv);
return model;

To reproduce

Create an image classification model with 4 channels and try to load the onnx model.

Urgency

Soon

Platform

Windows

OS Version

11

ONNX Runtime Installation

Released Package

ONNX Runtime Version or Commit ID

2.0.1

ONNX Runtime API

C#

Architecture

X64

Execution Provider

CUDA

Execution Provider Library Version

11.8

[gktval]

Simple mistake, I now see the option for rgba in extract pixels

mlContext.Transforms.ExtractPixels("sequential_input", "image", Microsoft.ML.Transforms.Image.ImagePixelExtractingEstimator.ColorBits.All)