example_inference.py

import torch
from PIL import Image
from torchvision import transforms
import timm, json

labels = [
  'crevice_corrosion',
  'erosion_corrosion',
  'galvanic_corrosion',
  'mic_corrosion',
  'no_corrosion',
  'pitting_corrosion',
  'stress_corrosion',
  'under_insulation_corrosion',
  'uniform_corrosion'
]

model = timm.create_model('resnet50', pretrained=False, num_classes=len(labels))
state = torch.load('resnet50-corrosion-classifier-v1.pth', map_location='cpu')
model.load_state_dict(state, strict=False)
model.eval()

transform = transforms.Compose([
    transforms.Resize(256, interpolation=transforms.InterpolationMode.BICUBIC),
    transforms.CenterCrop(224),
    transforms.ToTensor(),
    transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]),
])

def predict(path):
    img = Image.open(path).convert('RGB')
    x = transform(img).unsqueeze(0)
    with torch.no_grad():
        probs = model(x).softmax(dim=1).squeeze().tolist()
    idx = int(torch.tensor(probs).argmax())
    return labels[idx], probs[idx]

if __name__ == "__main__":
    import sys
    print(predict(sys.argv[1] if len(sys.argv)>1 else "test.jpg"))