How to infer on an image without saving it to a file

[lidiahhhh]

Hello,
I would like to infer an image that is in RAM without saving it to a file and then creating a Dataset on the image path.

Can I call predict directly on the image array? The image is the result of an object detection step.
Thanks

[tominator95]

Hi,
I created a REST API for inference. When I send the request, I send the file within it. So its not saved anywhere. The inferencer.predict() fucntion expects a path to an image file or the image as numpy array. More in the code snippet:

from anomalib.deploy import TorchInferencer
from PIL import Image

inferencer = TorchInferencer(config="<<path to config>>", model_source="<<path to weights file>>", device="auto")

def predict(file: typing.IO[bytes]):
        file_bytes = bytearray(file.read())
        img = Image.open(io.BytesIO(file_bytes))
        img_arr = np.array(img)
       
       predictions = inferencer.predict(image=img_arr)

I hope that helps you.

[lidiahhhh]

Thanks! It really helped me