how to inference model onnx with dynamic input sizes in ml.net ?

I want to infer model.onnx with animation size using ml.net or any other library on .net. I used ml.net and it seems to require declaring a fixed input size and the image must be resized to that size before being added to the model.