applications

# CLIP 应用指南 CLIP 的实际应用与使用案例。 ## 零样本图像分类 python import torch import clip from PIL import Image model, preprocess = clip.load("ViT-B/32") # 定义类别 categories = [ "a photo of a dog", "a photo of a cat", "a photo of a bird", "a photo of a car", "a photo of a person" ] # 准备图像 image = preprocess(Image.open("photo.jpg")).unsqueeze(0) text = clip.tokenize(categories) # 分类 with torch.no_grad(): image_features = model.encode_image(image) text_features = model.encode_text(text) logits_per_image, _ = model(image, text) probs = logits_per_image.softmax(dim=-1).cpu().numpy() # 打印结果 for category, prob in zip(categories, probs[0]): print(f"{category}: {prob:.2%}") ## 语义图像搜索 python # 索引图像 image_database = [] image_paths = ["img1.jpg", "img2.jpg", "img3.jpg"] for img_path in image_paths: image = preprocess(Image.open(img_path)).unsqueeze(0) with torch.no_grad(): features = model.encode_image(image) features /= features.norm(dim=-1, keepdim=True) image_database.append((img_path, features)) # 使用文本搜索 query = "a sunset over mountains" text_input = clip.tokenize([query]) with torch.no_grad(): text_features = model.encode_text(text_input) text_features /= text_features.norm(dim=-1, keepdim=True) # 查找匹配项 similarities = [] for img_path, img_features in image_database: similarity = (text_features @ img_features.T).item() similarities.append((img_path, similarity)) # 按相似度排序 similarities.sort(key=lambda x: x[1], reverse=True) for img_path, score in similarities[:3]: print(f"{img_path}: {score:.3f}") ## 内容审核 python # 定义安全类别 categories = [ "safe for work content", "not safe for work content", "violent or graphic content", "hate speech or offensive content", "spam or misleading content" ] text = clip.tokenize(categories) # 检查图像 with torch.no_grad(): logits, _ = model(image, text) probs = logits.softmax(dim=-1) # 获取分类结果 max_idx = probs.argmax().item() confidence = probs[0, max_idx].item() if confidence > 0.7: print(f"Classified as: {categories[max_idx]} ({confidence:.2%})") else: print(f"Uncertain classification (confidence: {confidence:.2%})") ## 图文检索 python # 文本数据库 captions = [ "A beaut