How to Load a Custom-Trained ResNet50 Model with ImageAI in Python?

I'm working on an image classification project using a custom-trained ResNet50 model with ImageAI, but I'm encountering a persistent loading error that prevents me from using my trained model for predictions. The error suggests there's a compatibility issue or a specific method I'm missing when trying to load a custom-trained PyTorch model into the ImageAI framework.

I attempted to load my custom-trained ResNet50 model using ImageAI's standard model loading procedure:

from imageai.Classification import ImageClassification

classifier = ImageClassification()
classifier.setModelPath("path/to/custom/model.pt")
classifier.setModelTypeAsResNet50()
classifier.loadModel()  # Expected to successfully load the model

What I expected was a straightforward model loading process similar to using pre-trained models. Instead, I received a frustrating error:

Classification failed: Model not yet loaded. You need to call '.loadModel()' 
before performing image classification

My environment details:

• Python 3.12
• ImageAI 3.0.3
• PyTorch 2.5.1+cu118
• Windows 11

I've confirmed:

• The model path is correct
• The model is saved as a valid .pt file
• The model was trained on a custom dataset using ResNet50 architecture

Specific Questions:

• Is there a specific method for loading custom-trained models in ImageAI?
• Are there any special export or conversion requirements for custom models?
• Do I need to modify my model's state dictionary or use a specific export format?