inference.py

import time
from PIL import Image
from timm.data import resolve_data_config
import torch
from torchvision.transforms import transforms

model = torch.load('path/to/model.pth')
model.eval()

config = resolve_data_config({}, model=model)
transform = transforms.Compose([
 transforms.Resize((224, 224)),
 transforms.ToTensor(),
 transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]) # Normalize image
 ])


with open("tags.txt", "r") as f:
 categories = [s.strip() for s in f.readlines()]
 categories=sorted(categories)


device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')

images=["your_image_here.jpg", "your_second_image_here.jpg"]

for item in images:
 start = time.time()
 img = Image.open(item).convert('RGB')
 tensor = transform(img).unsqueeze(0).to(device) # transform and add batch dimension

 with torch.no_grad():
 out = model(tensor)
 probabilities = torch.nn.functional.sigmoid(out[0])
 print(probabilities.shape)


 top10_prob, top10_catid = torch.topk(probabilities, 10)
 for i in range(top10_prob.size(0)):
 print(categories[top10_catid[i]], top10_prob[i].item())

 end = time.time()
 print(f'Executed in {end - start} seconds')
 print("\n\n", end="")