app.py

import gradio as gr
import tensorflow as tf
import numpy as np
from tensorflow.keras.models import load_model
from PIL import Image
import cv2
import numpy as np
import gradio as gr
import zipfile
from io import BytesIO
# ham_dict = {0:"nv:melanocytic nevi", 1:"mel:melanoma", 2:"bkl:benign keratosis-like lesions ", 3:"bcc:basal cell carcinoma ", 4:"akiec:Actinic keratoses and intraepithelial carcinoma / Bowen's disease ", 5:"vasc:vascular lesions", 6:"df:dermatofibroma"}
ham_dict = {0: "nv(黑素细胞痣)", 1: "mel(黑色素瘤)", 2: "bkl(良性角化病样病变)",
            3: "bcc(基底细胞癌)", 4: "akiec(日光性角化病和上皮内癌/博温病)",
            5: "vasc(血管性病变)", 6: "df(皮肤纤维瘤)"}
model = load_model('my_multi_model_2.h5')

# def flip_text(x):
#     return x[::-1]
def classify_image(image):
    image = image.resize((224, 224))
    image = np.array(image)
    image = np.expand_dims(image, axis=0) 
    predictions = model.predict(image)
    predicted_class = np.argmax(predictions)
    str = ham_dict[predicted_class]
    print(str)
    class_name = str.split('(')[0]
    if class_name == 'bkl' or class_name == 'nv':
        result = '良性'
    else:
        result = '恶性'
    return f" 皮肤信息：{str} 诊断结果：{result}"

# def flip_image(x):
#     return np.fliplr(x)

# 定义处理函数，该函数接受上传的ZIP文件并处理其中的图像
def process_zip_file(zip_file):
    image_messages = {}
    print(zip_file.name)
    print(type(zip_file))
    with zipfile.ZipFile(zip_file.name, "r") as zf:
        for file_name in zf.namelist():
            if file_name.lower().endswith(('.png', '.jpg')):
                with zf.open(file_name) as img_file:
                    img_data = img_file.read()
                    img = Image.open(BytesIO(img_data))
                    image_result = classify_image(img)
                    image_messages[file_name] = image_result
    if not image_messages:
        return "ZIP文件中没有图像文件"
    
    return "\n".join([f"{key.split('/')[-1]}: {value}" for key, value in image_messages.items()])

# a = process_zip_file('/gemini/code/皮肤.zip')
# print(a)
with gr.Blocks() as demo:
    gr.Markdown("皮肤癌检测,提示：该模型诊断结果置信度为0.8，只能作为在线参考，及时就医才是王道！")
    with gr.Tab("研究目的"):
        gr.Markdown("通过检测识别出皮肤癌的发病机制,可以更好地防治和治疗皮肤癌。视频介绍皮肤癌的常见类型和发病机制。")
        gr.Video("ham.mp4")
    with gr.Tab("单张图片检测（上传png,jpg图片）"):
        input_interface = gr.Image(type='pil', label="上传图片")
        output_interface = gr.Textbox(label="预测结果")
        text_button = gr.Button("单张检测")
    with gr.Tab("多张图片检测（上传zip文件）"):
        # with gr.Row():
        input_component = gr.File(label="上传ZIP文件") #file指定路径
        out_ = gr.Textbox(label="预测结果")
        zip_button = gr.Button("多张检测")

    with gr.Accordion("注意事项"):
        gr.Markdown("1.只支持上传zip压缩包 \n 2.压缩包中的图片名用英语或数字表示（中文图片名检测会出现乱码，不方便辨别）\n 3.服务器算力有限，多图片处理得比较慢，请耐心等待，或换成单图片处理")
        gr.Markdown('[我的网站](https://hpa888.top/)')
    text_button.click(classify_image, inputs=input_interface, outputs=output_interface)
    zip_button.click(process_zip_file, inputs=input_component, outputs=out_)
demo.launch()