What is the correct way to use the image output of LoadImage as input for mmdet.inference_detector?

[ltdrdata]

I am currently studying MMDetection in order to replicate the DDetailer project, and I have almost completed the basic setup.

I have been using the inference_detector function in MMDetection, and it seems to work perfectly when the image parameter is a filepath.

        mmdet_results = inference_detector(model, "/home/.../example.png")

However, if the input for the image parameter is a ndarray, the function execution fails during the resizing stage in the pipeline. I'm having trouble understanding the behavior of the pipeline, especially since there appear to be several transformations already in place.

        mmdet_results = inference_detector(model, image.numpy())

("Image" is obtained from the output of "LoadImage" node.)

According to the API description, the img parameter allows both filenames and ndarrays as input. However, I am experiencing errors when using an ndarray as input, and I am not familiar with the framework to understand why this is happening.

How to solve this problem?

[ltdrdata]

problem solved.

    image = image.numpy()[0] * 255

This is everything what I needed.