process/preprocess.py

import os
import torch
import json
import shutil

def is_connected(graph):
    # 获取所有的顶点
    nodes = list(graph.keys())

    # 如果图中没有节点，则认为是连通的（空图被视为连通）
    if not nodes:
        return False

    # 初始化一个集合用于记录访问过的节点
    visited = set()

    # 深度优先搜索 DFS
    def dfs(node):
        visited.add(node)
        for neighbor in graph[node]:
            if neighbor not in visited:
                dfs(neighbor)

    # 从第一个节点开始 DFS
    dfs(nodes[0])

    # 检查是否所有节点都被访问过
    return len(visited) == len(nodes)


folder_path = './dataset/parts_'
if not os.path.exists(folder_path):
    os.makedirs(folder_path)
cad_classes = ['building', 'chair', 'fan', 'lamp', 'table', 'tools', 'vehicle']
for cad_class in cad_classes:
    folder_name = f'./dataset/assemblies/{cad_class}'
    file_list = os.listdir(folder_name)
    file_list = sorted(file_list, key=lambda x: int(x.split('_')[1]))
    for file_name in file_list:
        full_name = os.path.join(folder_name, file_name)
        parts_list = os.listdir(full_name)
        parts_list = [part for part in parts_list if part.startswith('part')]
        parts_list = sorted(parts_list, key=lambda x: int(x.split('_')[1].split('.')[0]))
        graph_path = f'{full_name}/processed/data.pt'
        if not os.path.exists(graph_path):
            with open("./filtered/filtered_assembly.txt", "a") as f:
                f.write(f"{full_name}，图为空\n")
            print(f"{full_name}被过滤掉，图为空")
            # 删除文件夹及其内容
            if os.path.exists(full_name):
                shutil.rmtree(full_name)
                print(f"{full_name} 文件夹已被删除")
            continue
        data, slices = torch.load(graph_path)
        num_nodes = data['x'].size()[0]
        if num_nodes > 15:
            with open("./filtered/filtered_assembly.txt", "a") as f:
                f.write(f"{full_name}，零件数量超过15\n")
            print(f"{full_name}被过滤掉，零件数量超过15")
            # 删除文件夹及其内容
            if os.path.exists(full_name):
                shutil.rmtree(full_name)
                print(f"{full_name} 文件夹已被删除")
            continue
        edges = data['edge_index']
        if edges.size()[0] != 2:
            with open("./filtered/filtered_assembly.txt", "a") as f:
                f.write(f"{full_name}，图为空\n")
            print(f"{full_name}被过滤掉，图为空")
            # 删除文件夹及其内容
            if os.path.exists(full_name):
                shutil.rmtree(full_name)
                print(f"{full_name} 文件夹已被删除")
            continue
        graph = {i: [] for i in range(num_nodes)}
        for i in range(edges.size()[1]):
            u, v = edges[0][i].item(), edges[1][i].item()
            graph[u].append(v)
        if is_connected(graph):
            parts_all_list = os.listdir(folder_path)
            num_parts_all = len(parts_all_list)
            with open(f'{full_name}/original_graph.json', 'w') as f:
                json.dump(graph, f)
            new_graph = {}
            for i in range(len(parts_list)):
                renamed_file = os.path.join(full_name, f'parts_{i + num_parts_all + 809}.step')
                original_file = os.path.join(full_name, parts_list[i])
                os.rename(original_file, renamed_file)
                new_file_path = os.path.join(folder_path, os.path.basename(renamed_file))
                shutil.copy(renamed_file, new_file_path)
                new_graph[f'parts_{i + num_parts_all + 809}'] = [f'parts_{j + num_parts_all + 809}' for j in graph[i]]
            with open(f'{full_name}/new_graph.json', 'w') as f:
                json.dump(new_graph, f)
        else:
            with open("./filtered/filtered_assembly.txt", "a") as f:
                f.write(f"{full_name}，装配体不连通\n")
            print(f"{full_name}被过滤掉，装配体不连通")
            # 删除文件夹及其内容
            if os.path.exists(full_name):
                shutil.rmtree(full_name)
                print(f"{full_name} 文件夹已被删除")