app.py

from pathlib import Path
import torch
import gradio as gr
from torch import nn
import numpy as np

print(gr.__version__)

LABELS = Path("class_names.txt").read_text().splitlines()

model = nn.Sequential(
    nn.Conv2d(1, 32, 3, padding="same"),
    nn.ReLU(),
    nn.MaxPool2d(2),
    nn.Conv2d(32, 64, 3, padding="same"),
    nn.ReLU(),
    nn.MaxPool2d(2),
    nn.Conv2d(64, 128, 3, padding="same"),
    nn.ReLU(),
    nn.MaxPool2d(2),
    nn.Flatten(),
    nn.Linear(1152, 256),
    nn.ReLU(),
    nn.Linear(256, len(LABELS)),
)

state_dict = torch.load("pytorch_model.bin", map_location="cpu")
model.load_state_dict(state_dict, strict=False)
model.eval()

def predict(im):
    if im is None:
        return {}
    
    # 处理输入图像
    # 如果是字典格式（新版Gradio Sketchpad的输出），提取图像
    if isinstance(im, dict):
        im = im['image'] if 'image' in im else im.get('composite', None)
    
    # 转换为numpy数组并确保是灰度图
    if isinstance(im, np.ndarray):
        if len(im.shape) == 3:
            # 如果是RGB图像，转换为灰度图
            im = np.mean(im, axis=2)
    else:
        return {}
    
    # 确保图像尺寸正确（28x28）
    if im.shape != (28, 28):
        from PIL import Image
        im_pil = Image.fromarray(im.astype('uint8'))
        im_pil = im_pil.resize((28, 28))
        im = np.array(im_pil)
    
    # 转换为tensor并进行预测
    x = torch.tensor(im, dtype=torch.float32).unsqueeze(0).unsqueeze(0) / 255.0
    
    with torch.no_grad():
        out = model(x)
    
    probabilities = torch.nn.functional.softmax(out[0], dim=0)
    values, indices = torch.topk(probabilities, 5)
    
    return {LABELS[i]: v.item() for i, v in zip(indices, values)}

# 创建Gradio界面
interface = gr.Interface(
    fn=predict,
    inputs=gr.Sketchpad(
        image_mode="L",  # 灰度模式
        canvas_size=(280, 280),  # 画布大小
        brush=gr.Brush(default_size=10)  # 画笔设置
    ),
    outputs=gr.Label(num_top_classes=5),  # 显示前5个预测结果
    title="Sketch Recognition",
    description="Who wants to play Pictionary? Draw a common object like a shovel or a laptop, and the algorithm will guess in real time!",
    article="<p style='text-align: center'>Sketch Recognition | Demo Model</p>",
)

interface.launch(share=True)