app.py

import gradio as gr
import cv2
import tempfile
from ultralytics import YOLO
from pathlib import Path

# 全局变量存储当前模型
current_model = None

def load_model(model_path):
    global current_model
    try:
        current_model = YOLO(model_path)
        return "模型加载成功！"
    except Exception as e:
        return f"模型加载失败：{str(e)}"

def detect_image(input_image, conf_threshold):
    if current_model is None:
        raise gr.Error("请先上传模型文件")
    
    results = current_model(input_image, conf=conf_threshold)
    plotted = results[0].plot()
    return plotted[:, :, ::-1]  # BGR转RGB

def detect_video(input_video, conf_threshold):
    if current_model is None:
        raise gr.Error("请先上传模型文件")
    
    cap = cv2.VideoCapture(input_video)
    fps = cap.get(cv2.CAP_PROP_FPS)
    width = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH))
    height = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))
    
    # 创建临时输出文件
    temp_file = tempfile.NamedTemporaryFile(suffix=".mp4", delete=False)
    out = cv2.VideoWriter(temp_file.name, cv2.VideoWriter_fourcc(*'mp4v'), fps, (width, height))
    
    while cap.isOpened():
        ret, frame = cap.read()
        if not ret:
            break
        
        results = current_model(frame, conf=conf_threshold)
        plotted = results[0].plot()
        out.write(plotted)
    
    cap.release()
    out.release()
    return temp_file.name

def detect_webcam(camera_input, conf_threshold):
    if current_model is None:
        raise gr.Error("请先上传模型文件")
    
    if camera_input is None:
        return None
    
    results = current_model(camera_input, conf=conf_threshold)
    plotted = results[0].plot()
    return plotted[:, :, ::-1]  # BGR转RGB

with gr.Blocks() as demo:
    gr.Markdown("# YOLOv8 自定义模型检测系统")
    
    with gr.Row():
        model_input = gr.File(label="上传模型文件 (.pt)", type="filepath")
        model_status = gr.Textbox(label="模型状态", interactive=False)
    
    model_input.upload(fn=load_model, inputs=model_input, outputs=model_status)
    
    with gr.Tabs():
        with gr.TabItem("图片检测"):
            with gr.Row():
                img_input = gr.Image(label="输入图片", type="filepath")
                img_output = gr.Image(label="检测结果")
            img_conf = gr.Slider(0, 1, value=0.5, label="置信度阈值")
            img_button = gr.Button("执行检测")
        
        with gr.TabItem("视频检测"):
            with gr.Row():
                video_input = gr.Video(label="输入视频")
                video_output = gr.Video(label="检测结果")
            video_conf = gr.Slider(0, 1, value=0.5, label="置信度阈值")
            video_button = gr.Button("执行检测")
        
        with gr.TabItem("实时摄像头"):
            webcam_input = gr.Webcam(label="摄像头画面")  # 使用官方 Webcam 组件
            webcam_output = gr.Image(label="检测结果")
            webcam_conf = gr.Slider(0, 1, value=0.5, label="置信度阈值")
            webcam_button = gr.Button("开始检测")

webcam_button.click(fn=detect_webcam, inputs=[webcam_input, webcam_conf], outputs=webcam_output)
    
    # 绑定事件处理
    img_button.click(fn=detect_image, inputs=[img_input, img_conf], outputs=img_output)
    video_button.click(fn=detect_video, inputs=[video_input, video_conf], outputs=video_output)
    webcam_button.click(fn=detect_webcam, inputs=[webcam_input, webcam_conf], outputs=webcam_output)

if __name__ == "__main__":
    demo.launch()