atom-predict/msunet/streamlit_app.py

import streamlit as st
import os
import glob
import shutil
import argparse
import os
import zipfile
import glob
from typing import List
from pre import process_slide
from test_pl_unequal import predict
from typing import Dict
from typing import Tuple
from utils.labelme import save_pred_to_json
from plot import plot2
from predict import predict_and_plot
from train_pl import train, pre_process, create_save_path
import numpy as np
from dp.launching.report import Report, ReportSection, AutoReportElement
import sys
from pathlib import Path


sys.path.append(os.path.abspath(os.path.join(os.path.dirname(__file__), '../egnn_v2')))
print(sys.path)
from predict_pl_v2_aug import predict_and_plot as egnn_predict_and_plot, gnn_generate_report
from train_pl_v2_aug import train as egnn_train

# 获取运行脚本目录中的 `model` 文件夹内的所有 `*.ckpt` 文件
def get_model_options(model_dir="../egnn_v2/model"):
    ckpt_files = glob.glob(os.path.join(model_dir, "*.ckpt"))
    model_options = [os.path.basename(f).replace(".ckpt", "") for f in ckpt_files]
    return model_options

def load_images_from_folder(folder_path):
    images = []
    for filename in os.listdir(folder_path):
        if filename.endswith(('.png', '.jpg', '.jpeg')):
            img_path = os.path.join(folder_path, filename)
            img = Image.open(img_path)
            images.append((filename, img))
    return images


# 初始化
if 'models' not in st.session_state:
    st.session_state['models'] = []

# 页面标题
st.title("Bohrium Apps 开发者中心")

# 创建 Tabs
tabs = st.tabs(["combine_gnn预测", "新增模型"])

def run_task(data_file,select_models, edges_length):
    # 打印调试信息
    st.write(f"Edges Length: {edges_length}")
    st.write(f"Selected Models: {select_models}")
    st.write(f"data_file: {data_file.name}")
    # 保存zip文件到临时目录, 临时目录为当前目录下的 `temp` 文件夹，使用绝对路径
    base_dir = os.path.join(os.path.dirname(os.path.abspath(__file__)))

    temp_dir = os.path.join(base_dir, "temp")
    os.makedirs(temp_dir, exist_ok=True)
    temp_file_path = os.path.join(temp_dir, data_file.name)


    with open(temp_file_path, "wb") as f:
        f.write(data_file.getbuffer())

    print("select_models的类型是：", type(select_models))
    print("选择的模型:",select_models)
    for model in select_models:
        print("选择的模型的名称:",model)


    output_dir = "./streamlit_app_result"
    model_path = os.path.join(base_dir, "model/last.ckpt")
    save_path = predict_and_plot(model_path, temp_file_path, output_dir)

    test_path = save_path["post_process"]
    print("gnn 预测开始")

    for opt in select_models:
        gnn_model_path = os.path.join(os.path.dirname(os.path.abspath(__file__)), "../egnn_v2/model", opt+".ckpt")
        print("gnn model path :", gnn_model_path)
        egnn_predict_and_plot(gnn_model_path, test_path, output_dir, edges_length)

    print("gnn 预测结束")

    return {
        "result_view": os.path.join(output_dir, "gnn_predict_result_view"),
        "connect_view": os.path.join(output_dir, "gnn_predict_connect_view")
    }

# 用于展示结果的缓存
result_paths = {
    "result_view": None,
    "connect_view": None
}


with tabs[0]:
   # st.header("combine_gnn预测")

    # Data Path
    data_file = st.file_uploader("Data Path", type=["zip"], help="测试的缺陷数据集")

    # Model Path
    #model_file = st.file_uploader("Model Path", type=["ckpt"], help="原子缺陷处理模型")

    # Gnn Model Path
    #gnn_model_file = st.file_uploader("Gnn Model Path", type=["ckpt"], help="gnn模型")

    # Select Models
  #  st.subheader("Select Models")

    # 获取模型选项
    model_options = get_model_options()

    # 多选框
    selected_models = st.multiselect("选择模型", model_options, default=st.session_state['models'])
    st.session_state['models'] = selected_models

    # Edges Length
    edges_length = st.number_input("Edges Length", value=35.5)

    # 运行任务按钮
    if st.button("运行任务"):
        if not data_file :
            st.error("请上传所有必需的文件")
        else:
            # 模拟任务运行
            st.success("任务运行中...")
            # 获取选中的模型
            select_models = st.session_state['models']
            # 获取边长

            run_task(data_file, select_models, edges_length)
            st.success("任务完成")
            # 展示结果
            if result_paths["result_view"]:
                st.subheader("GNN 预测结果")
                result_images = load_images_from_folder(result_paths["result_view"])
                for filename, img in result_images:
                    st.image(img, caption=filename, use_column_width=True)

            if result_paths["connect_view"]:
                st.subheader("连接视图")
                connect_view_images = load_images_from_folder(result_paths["connect_view"])
                for filename, img in connect_view_images:
                    st.image(img, caption=filename, use_column_width=True)

with tabs[1]:
    st.header("新增模型")

    # 上传模型文件
    new_model_file = st.file_uploader("Model Path", type=["ckpt"], help="上传新的模型文件 (.ckpt)")

    if st.button("提交"):
        if new_model_file is None:
            st.error("请上传模型文件")
        else:
            model_dir = "../egnn_v2/model"
            os.makedirs(model_dir, exist_ok=True)
            model_file_path = os.path.join(model_dir, new_model_file.name)

            # 保存上传的模型文件
            with open(model_file_path, "wb") as f:
                f.write(new_model_file.getbuffer())

            st.success(f"模型文件已保存到 {model_file_path}")
            st.rerun()  # 刷新页面以更新模型选项