# app.py (Gradio界面) import gradio as gr import requests from PIL import Image import numpy as np from sklearn.cluster import KMeans import time import random import os import torch # 模型管理器 from models.model_manager import ModelManager # 初始化模型管理器 model_manager = ModelManager() def upload_and_analyze(image_path): """分析上传的图片""" try: if image_path is None: return {}, {}, [] # 打开图片 image = Image.open(image_path).convert('RGB') # 生成图像描述 caption = model_manager.generate_caption(image) # 基于图像描述进行智能分析 analysis_result = analyze_image_content(image, caption) # 基于分析结果生成个性化建议 suggestions = generate_personalized_suggestions(analysis_result, caption) # 创建选择选项 choices = list(suggestions.keys()) return analysis_result, suggestions, gr.Radio(choices=choices, value=choices[0] if choices else None) except Exception as e: error_result = {"错误": f"分析失败: {str(e)}"} return error_result, {}, [] def analyze_image_content(image, caption): """基于图像和描述进行深度分析""" # 分析图像颜色 colors = extract_dominant_colors(image) # 根据描述推断风格类型 style_type = infer_style_from_caption(caption) # 根据描述推断服装类别 clothing_category = infer_clothing_category(caption) # 根据风格推荐适合场景 suitable_scenes = get_suitable_scenes(style_type) return { "图像描述": caption, "检测到的颜色": colors, "风格类型": style_type, "服装类别": clothing_category, "适合场景": suitable_scenes, "图像尺寸": f"{image.width} x {image.height}", "分析时间": time.strftime("%Y-%m-%d %H:%M:%S") } def extract_dominant_colors(image): """提取图像主要颜色""" # 调整图像大小以提高处理速度 image = image.resize((150, 150)) # 转换为numpy数组 img_array = np.array(image) # 重塑为颜色列表 pixels = img_array.reshape(-1, 3) # 使用KMeans聚类找到主要颜色 kmeans = KMeans(n_clusters=3, random_state=42, n_init=10) kmeans.fit(pixels) # 将RGB值转换为颜色名称 color_names = [] for color in kmeans.cluster_centers_: color_name = rgb_to_color_name(color) color_names.append(color_name) return color_names def rgb_to_color_name(rgb): """将RGB值转换为颜色名称""" r, g, b = rgb.astype(int) # 简单的颜色映射 if r > 200 and g > 200 and b > 200: return "白色" elif r < 50 and g < 50 and b < 50: return "黑色" elif r > g and r > b: if r > 150: return "红色" else: return "深红色" elif g > r and g > b: if g > 150: return "绿色" else: return "深绿色" elif b > r and b > g: if b > 150: return "蓝色" else: return "深蓝色" elif r > 150 and g > 150: return "黄色" elif r > 100 and b > 100: return "紫色" elif g > 100 and b > 100: return "青色" else: return "灰色" def infer_style_from_caption(caption): """根据图像描述推断风格类型""" caption_lower = caption.lower() style_keywords = { "商务正装": ["suit", "formal", "business", "office", "professional", "tie", "blazer", "西装", "正装", "商务"], "休闲风": ["casual", "relaxed", "comfortable", "everyday", "jeans", "t-shirt", "休闲", "日常"], "运动风": ["sport", "athletic", "gym", "fitness", "running", "training", "运动", "健身"], "时尚潮流": ["fashion", "trendy", "stylish", "modern", "chic", "designer", "时尚", "潮流"], "复古风": ["vintage", "retro", "classic", "traditional", "old-fashioned", "复古", "经典"], "街头风": ["street", "urban", "hip-hop", "cool", "edgy", "街头", "嘻哈"], "优雅风": ["elegant", "sophisticated", "graceful", "refined", "classy", "优雅", "高贵"] } for style, keywords in style_keywords.items(): if any(keyword in caption_lower for keyword in keywords): return style return "休闲风" # 默认风格 def infer_clothing_category(caption): """根据描述推断服装类别""" caption_lower = caption.lower() categories = { "上衣": ["shirt", "blouse", "top", "jacket", "sweater", "hoodie", "blazer", "衬衫", "上衣", "外套"], "下装": ["pants", "jeans", "skirt", "shorts", "trousers", "裤子", "短裤", "裙子"], "连衣裙": ["dress", "gown", "frock", "连衣裙", "礼服"], "外套": ["coat", "jacket", "cardigan", "blazer", "外套", "大衣"], "配饰": ["hat", "bag", "shoes", "belt", "jewelry", "帽子", "包", "鞋子", "配饰"], "全套搭配": ["outfit", "ensemble", "look", "style", "搭配", "整套"] } for category, keywords in categories.items(): if any(keyword in caption_lower for keyword in keywords): return category return "服装单品" # 默认类别 def get_suitable_scenes(style_type): """根据风格类型推荐适合场景""" scene_mapping = { "商务正装": ["办公室", "商务会议", "正式场合", "面试"], "休闲风": ["日常出街", "朋友聚会", "购物", "咖啡约会"], "运动风": ["健身房", "运动", "户外活动", "晨跑"], "时尚潮流": ["时尚派对", "约会", "拍照", "社交活动"], "复古风": ["艺术展", "文艺活动", "复古主题活动", "拍摄"], "街头风": ["街拍", "音乐节", "朋友聚会", "潮流活动"], "优雅风": ["晚宴", "正式聚会", "典礼", "高端场所"] } return scene_mapping.get(style_type, ["日常", "休闲", "约会"]) def generate_personalized_suggestions(analysis_result, caption): """基于分析结果生成个性化建议""" style_type = analysis_result["风格类型"] clothing_category = analysis_result["服装类别"] colors = analysis_result["检测到的颜色"] suggestions = {} # 根据检测到的风格生成建议 if style_type == "商务正装": suggestions = { "经典商务": f"保持{style_type}特色，搭配{colors[0]}系配饰", "现代商务": f"在{style_type}基础上加入现代元素", "休闲商务": f"将{style_type}与休闲元素结合", "时尚商务": f"{style_type}融入时尚潮流元素" } elif style_type == "休闲风": suggestions = { "舒适休闲": f"强化{style_type}的舒适感，主色调{colors[0]}", "时尚休闲": f"{style_type}加入时尚元素", "运动休闲": f"{style_type}融入运动风格", "优雅休闲": f"{style_type}提升优雅感" } elif style_type == "运动风": suggestions = { "专业运动": f"增强{style_type}的功能性", "休闲运动": f"{style_type}与日常穿着结合", "时尚运动": f"{style_type}加入潮流设计元素", "户外运动": f"强化{style_type}的户外适应性" } else: suggestions = { f"经典{style_type}": f"保持原有{style_type}特色", f"现代{style_type}": f"{style_type}加入现代元素", f"融合风格": f"{style_type}与其他风格混搭", f"个性化{style_type}": f"基于{colors[0]}色调的个性化{style_type}" } return suggestions def generate_designs(selected_suggestion, progress=gr.Progress()): """根据选择的建议生成设计""" try: if not selected_suggestion: return [], gr.Radio(choices=[]) progress(0.1, desc="准备设计提示...") # 生成设计图像的提示词 design_prompts = { "经典商务": "professional business suit, modern cut, high-quality fabric, clean lines, neutral colors", "现代商务": "contemporary business attire, innovative design, slim fit, premium materials", "休闲商务": "business casual outfit, comfortable yet professional, versatile style", "时尚商务": "fashion-forward business wear, runway inspired, statement piece", "舒适休闲": "casual comfort wear, soft fabrics, relaxed fit, everyday style", "时尚休闲": "stylish casual outfit, trendy elements, urban chic", "运动休闲": "athleisure wear, sporty elements, comfortable and functional", "优雅休闲": "elegant casual attire, sophisticated details, refined look", "专业运动": "performance sportswear, technical fabrics, functional design", "休闲运动": "casual athletic wear, versatile for sports and daily use", "时尚运动": "fashion sportswear, trendy athletic style, streetwear influence", "户外运动": "outdoor adventure wear, durable materials, weather-resistant", "经典复古风": "vintage retro style, classic silhouette, nostalgic elements", "现代复古风": "contemporary take on retro fashion, updated classics", "融合风格": "fusion fashion, mixed styles, innovative combination", "个性化街头风": "personalized streetwear, unique designs, urban style", "经典优雅风": "timeless elegant fashion, sophisticated details, refined look", "现代优雅风": "modern elegant attire, contemporary sophistication" } prompt = design_prompts.get(selected_suggestion, "fashion design, stylish clothing") # 生成设计图像 design_images = [] design_choices = [] for i in range(3): # 生成3个设计 try: progress(0.2 + i*0.25, desc=f"生成设计方案 {i+1}/3...") # 生成真实的设计图像 image = model_manager.generate_image( prompt=f"{prompt}, design {i+1}, high detail, fashion illustration", negative_prompt="blurry, low quality, distorted, text, watermark", num_inference_steps=30 ) if image: design_images.append(image) design_choices.append(f"{selected_suggestion} 设计方案 {i+1}") except Exception as e: print(f"生成设计 {i+1} 失败: {e}") # 创建占位图像 width, height = 512, 512 img = Image.new('RGB', (width, height), color=(random.randint(0, 255), random.randint(0, 255), random.randint(0, 255))) design_images.append(img) design_choices.append(f"{selected_suggestion} 设计方案 {i+1}") progress(0.95, desc="完成设计生成") return design_images, gr.Radio(choices=design_choices, value=design_choices[0] if design_choices else None) except Exception as e: print(f"设计生成错误: {e}") return [], gr.Radio(choices=[]) def generate_3d_fitting(selected_design, progress=gr.Progress()): """生成3D试穿效果""" try: if not selected_design: return None progress(0.1, desc="准备3D试穿...") # 生成3D试穿效果的提示词 fitting_prompt = f"3D fashion fitting, virtual try-on, {selected_design}, realistic human model, full body, studio lighting" progress(0.3, desc="生成3D模型...") # 使用模型生成3D试穿图像 if model_manager.controlnet_pipeline: # 如果有ControlNet，使用更高级的生成 try: # 这里简化了ControlNet的使用，实际需要姿势图等输入 image = model_manager.controlnet_pipeline( prompt=fitting_prompt, negative_prompt="blurry, distorted, low quality, unrealistic, extra limbs", num_inference_steps=35, guidance_scale=8.0 ).images[0] progress(0.9, desc="渲染3D效果") return image except Exception as e: print(f"使用ControlNet生成失败: {e}") # 回退到普通SD模型 progress(0.4, desc="使用标准模型生成...") image = model_manager.generate_image( prompt=fitting_prompt, negative_prompt="blurry, distorted, low quality, unrealistic, extra limbs", num_inference_steps=35 ) progress(0.9, desc="完成3D渲染") return image except Exception as e: print(f"3D试穿生成错误: {e}") return None def create_gradio_interface(): """创建Gradio用户界面""" with gr.Blocks(title="AI时尚设计师", theme="soft") as demo: gr.Markdown("# 🎨 AI时尚设计师") gr.Markdown("上传图片，获得专业的服装设计建议和3D试穿效果") with gr.Row(): with gr.Column(scale=1): image_input = gr.Image(type="filepath", label="上传参考图片", height=400) analyze_btn = gr.Button("分析风格", variant="primary") with gr.Column(scale=2): analysis_output = gr.JSON(label="风格分析结果") with gr.Tab("设计建议"): suggestions_output = gr.JSON(label="个性化设计建议") suggestion_choice = gr.Radio(label="选择设计方向", interactive=True) generate_designs_btn = gr.Button("生成设计方案", variant="primary") with gr.Tab("样衣设计"): designs_gallery = gr.Gallery(label="样衣设计图", columns=3, height=400) design_choice = gr.Radio(label="选择设计方案", interactive=True) generate_3d_btn = gr.Button("生成3D试穿效果", variant="primary") with gr.Tab("3D试穿效果"): fitting_result = gr.Image(label="3D试穿效果", height=500) # 添加示例图片 examples_dir = "examples" if os.path.exists(examples_dir): example_files = [os.path.join(examples_dir, f) for f in os.listdir(examples_dir) if f.lower().endswith(('.png', '.jpg', '.jpeg'))] gr.Examples( examples=example_files[:4], # 最多显示4个示例 inputs=image_input, outputs=[analysis_output, suggestions_output, suggestion_choice], fn=upload_and_analyze, cache_examples=True, label="示例图片" ) # 事件绑定 analyze_btn.click( fn=upload_and_analyze, inputs=[image_input], outputs=[analysis_output, suggestions_output, suggestion_choice] ) generate_designs_btn.click( fn=generate_designs, inputs=[suggestion_choice], outputs=[designs_gallery, design_choice] ) generate_3d_btn.click( fn=generate_3d_fitting, inputs=[design_choice], outputs=[fitting_result] ) # 添加清理按钮 clear_btn = gr.Button("清理内存", variant="secondary") clear_btn.click( fn=model_manager.cleanup, inputs=[], outputs=[] ) gr.Markdown("> **提示**: 生成图像后，点击'清理内存'按钮可以释放GPU资源") return demo if __name__ == "__main__": # 检查并创建示例目录 examples_dir = "examples" if not os.path.exists(examples_dir): os.makedirs(examples_dir) print(f"创建了示例目录: {examples_dir}") print("请在此目录中添加示例图片以便在界面中使用") demo = create_gradio_interface() demo.queue(concurrency_count=1) # 限制并发以避免内存问题 demo.launch( server_name="0.0.0.0", server_port=7860, share=True, favicon_path="favicon.ico" if os.path.exists("favicon.ico") else None )