[Bug]: OpenVINOInferencer does not auto resize the images leading to wrong inference results

[FedericoDeBona]

Describe the bug

I'm using a custom dataset with the same structure of MVTec datasets. When inferencing with OpenVINO, if the image is not resized to the train image size, the prediction will give wrong results. I think the same problem is also #2136

Wrong prediction

Inferenced with inferencer.predict(image=image_path) where the images are in 1440x1440

in this dataset images are 1024x1024

Correct prediction

Inferenced with

image = Image.open(img_path)
image = image.resize((256, 256))
predictions = inferencer.predict(image=image)

Torch inferencer

Using Torch inferencer the result is correct without resizing the images.

Code used for training:

datamodule = MVTec(
    category=CATEGORY,
    root=f"{ROOT}/datasets_original",
    image_size=(256,256))
model = Padim()
engine = Engine(default_root_dir=RESULTS_DIR)
engine.fit(datamodule=datamodule, model=model)
engine.export(
    model=model,
    export_type=ExportType.OPENVINO,
    compression_type=CompressionType.INT8
)
inferencer = OpenVINOInferencer(
    path=f"{RESULTS_DIR}/{MODEL}/MVTec/{CATEGORY}/latest/weights/openvino/model.bin",
    metadata=f"{RESULTS_DIR}/{MODEL}/MVTec/{CATEGORY}/latest/weights/openvino/metadata.json",
    device="CPU")

Dataset

Other (please specify in the text field below)

Model

PADiM

Steps to reproduce the behavior

see above

OS information

OS information:

• OS: Ubuntu 24
• Python version: 3.10.14
• Anomalib version: 1.2.0dev
• PyTorch version: 2.4.0+cu118
• CUDA/cuDNN version: 11.8
• GPU models and configuration: GeForce RTX 3090 Ti
  I'm using a custom dataset

Expected behavior

OpenVINOInferencer predicts correctly without resizing the images

Screenshots

Pip/GitHub

GitHub

What version/branch did you use?

No response

Configuration YAML

-

Logs

-

Code of Conduct

• I agree to follow this project's Code of Conduct

[watertianyi]

@FedericoDeBona

I used patchcore to convert OPENVINO, and the following error occurred when using cpu: image = cv2.resize(image, tuple(list(self.input_blob.shape)[2:][::-1]))
RuntimeError: Exception from src/core/src/partial_shape.cpp:266:
to_shape was called on a dynamic shape. Do you know why? I also get an error when using GPU. Can I not use NVIDIA GPU?