app.py

import torch
import gradio as gr
from ultralytics import YOLO
from PIL import Image

# 正确方式：信任 YOLOv8 的 DetectionModel 类（适配 PyTorch 2.6+）
torch.serialization.add_safe_globals(['ultralytics.nn.tasks.DetectionModel'])

# 加载模型
model = YOLO("best.pt")

# 定义推理函数
def predict(img):
    results = model(img)[0]
    output = []
    for box in results.boxes:
        cls_id = int(box.cls[0])
        label = results.names[cls_id]
        conf = float(box.conf[0])
        output.append(f"检测到：{label}（置信度：{conf*100:.1f}%）")
    return output  # ✅ 返回字符串列表，前端可读！

# 创建 Gradio 接口
iface = gr.Interface(
    fn=predict,
    inputs=gr.Image(type="pil"),
    outputs="json",
    title="眼部疾病检测（YOLOv8）",
    description="上传或拍摄眼部图像，模型将自动分析潜在的眼病。"
)

iface.launch()