| import json | |
| import math | |
| import time | |
| from PIL import Image | |
| import torch | |
| from torchvision.transforms import transforms | |
| device = torch.device('cuda' if torch.cuda.is_available() else 'cpu') | |
| model = torch.load("path/to/your/model.pth") | |
| model.to(device) | |
| model.eval() | |
| transform = transforms.Compose([ | |
| transforms.ToTensor(), | |
| transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]), | |
| ]) | |
| with open("tags_8034.json", "r") as f: | |
| tags = json.load(f) | |
| tags.append("placeholder0") | |
| tags = sorted(tags) | |
| image_path = "path/to/your/image.jpg" | |
| start = time.time() | |
| img = Image.open(image_path).convert('RGB') | |
| aspect_ratio = img.width / img.height | |
| new_height = math.sqrt(512 ** 2 / aspect_ratio) | |
| new_width = aspect_ratio * new_height | |
| img.thumbnail((int(new_width), int(new_height)), Image.LANCZOS) | |
| tensor = transform(img).unsqueeze(0).to(device) | |
| with torch.no_grad(): | |
| out = model(tensor) | |
| probabilities = torch.nn.functional.sigmoid(out[0]) | |
| indices = torch.where(probabilities > 0.3)[0] | |
| values = probabilities[indices] | |
| for i in range(indices.size(0)): | |
| print(tags[indices[i]], values[i].item()) | |
| end = time.time() | |
| print(f'Executed in {end - start} seconds') | |
| print("\n\n", end="") | |