You can not select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
399 lines
15 KiB
399 lines
15 KiB
|
|
import oss2 |
|
import os |
|
from tqdm import tqdm |
|
import os |
|
from pathlib import Path |
|
import numpy as np |
|
import os |
|
import argparse |
|
|
|
import open3d as o3d |
|
|
|
def custom_mesh_transform(vertices, transform_matrix): |
|
""" |
|
手动实现网格变换:对每个顶点应用齐次变换矩阵 |
|
参数: |
|
vertices: 网格顶点数组 (N, 3) |
|
transform_matrix: 4x4 齐次变换矩阵 |
|
返回: |
|
变换后的顶点数组 (N, 3) |
|
""" |
|
# 1. 顶点转齐次坐标 (N, 3) → (N, 4) |
|
homogeneous_vertices = np.hstack((vertices, np.ones((vertices.shape[0], 1)))) |
|
|
|
# 2. 应用变换矩阵:矩阵乘法 (4x4) * (4xN) → (4xN) |
|
transformed_homogeneous = transform_matrix @ homogeneous_vertices.T |
|
|
|
# 3. 转回非齐次坐标 (3xN) → (N, 3) |
|
transformed_vertices = transformed_homogeneous[:3, :].T |
|
return transformed_vertices |
|
|
|
class DataTransfer: |
|
''' |
|
数据传输类 |
|
''' |
|
def __init__(self, local_path: str, oss_path: str, oss_client: oss2.Bucket): |
|
''' |
|
local_path: 本地输出路径 |
|
oss_path: oss路径 |
|
oss_client: oss客户端 |
|
''' |
|
self.local_path = local_path |
|
self.oss_path = oss_path.lstrip('/') |
|
self.oss_client = oss_client |
|
|
|
order_id: str |
|
pid: str |
|
model_height: str |
|
|
|
def download_data_rename_json(self, json_model_info): |
|
""" |
|
从 OSS 下载数据到本地,保持原有目录结构 |
|
""" |
|
|
|
# 列出所有对象 |
|
objects = [] |
|
prefix = self.oss_path.lstrip('/') # 移除开头的 '/' 以匹配 OSS 格式 |
|
ishaveCodeJpg = False |
|
for obj in oss2.ObjectIterator(self.oss_client, prefix=prefix): |
|
if obj.key != prefix: # 跳过目录本身 |
|
print(f"obj.key={obj.key}") |
|
if ".jpg" in obj.key: |
|
if obj.key.endswith(f"printId_{json_model_info.print_id}Tex1.jpg"): |
|
objects.append(obj.key) |
|
ishaveCodeJpg = True |
|
else: |
|
objects.append(obj.key) |
|
|
|
if not ishaveCodeJpg: |
|
for obj in oss2.ObjectIterator(self.oss_client, prefix=prefix): |
|
if obj.key == f"{json_model_info.pid}Tex1.jpg": |
|
objects.append(obj.key) |
|
break |
|
|
|
print(f"objects={objects}") |
|
download_count = 0 |
|
# 下载所有文件,添加进度条 |
|
for obj_key in tqdm(objects, desc="下载进度"): |
|
if obj_key.endswith('/'): |
|
continue |
|
|
|
# if "printId" in obj_key: |
|
# continue |
|
|
|
# 计算相对路径 |
|
rel_path = obj_key[len(prefix):].lstrip('/') |
|
|
|
file_dir, file_name = os.path.split(rel_path) |
|
file_base, file_ext = os.path.splitext(file_name) |
|
|
|
# 根据文件后缀名进行重命名 |
|
if file_ext.lower() in ['.mtl', '.jpg', '.jpeg', '.png']: |
|
if file_name.endswith(f"printId_{json_model_info.print_id}Tex1.jpg"): |
|
file_name = file_name.replace(f"printId_{json_model_info.print_id}Tex1.jpg", f"{json_model_info.pid}Tex1.jpg") |
|
|
|
# 对于.mtl和图片文件,在原名前加order_id |
|
new_file_name = f"{json_model_info.order_id}_{file_name}" |
|
# new_file_name = file_name |
|
elif file_ext.lower() == '.obj': |
|
# 对于.obj文件,完全重命名 |
|
new_file_name = f"{json_model_info.obj_name}" |
|
else: |
|
# 其他文件类型保持原名 |
|
new_file_name = file_name |
|
print("new_file_name=", new_file_name) |
|
|
|
# 构建新的相对路径 |
|
if file_dir: # 如果有子目录 |
|
new_rel_path = os.path.join(file_dir, new_file_name) |
|
else: |
|
new_rel_path = new_file_name |
|
|
|
# 构建本地完整路径 |
|
local_path = os.path.join(self.local_path, new_rel_path) |
|
|
|
# 创建必要的目录 |
|
os.makedirs(os.path.dirname(local_path), exist_ok=True) |
|
|
|
# 下载文件 |
|
self.oss_client.get_object_to_file(obj_key, local_path) |
|
download_count += 1 |
|
if file_ext == '.obj': # 10MB以上 |
|
try: |
|
# 使用临时文件避免内存问题 [8](@ref) |
|
temp_path = local_path + '.tmp' |
|
with open(local_path, 'r', encoding='utf-8') as f_in, \ |
|
open(temp_path, 'w', encoding='utf-8') as f_out: |
|
|
|
mtllib_modified = False |
|
for line in f_in: |
|
if not mtllib_modified and line.strip().startswith('mtllib '): |
|
parts = line.split(' ', 1) |
|
if len(parts) > 1: |
|
old_mtl_name = parts[1].strip() |
|
new_mtl_name = f"{json_model_info.order_id}_{old_mtl_name}" |
|
f_out.write(f"mtllib {new_mtl_name}\n") |
|
mtllib_modified = True |
|
continue |
|
f_out.write(line) |
|
|
|
os.replace(temp_path, local_path) # 原子性替换 |
|
|
|
except IOError as e: |
|
print(f"处理大文件 {local_path} 时出错: {e}") |
|
if os.path.exists(temp_path): |
|
os.remove(temp_path) |
|
|
|
# 优化后的.obj文件处理逻辑 |
|
if file_ext == '.mtl': |
|
try: |
|
# 使用更高效的文件读取方式 [6,8](@ref) |
|
with open(local_path, 'r', encoding='utf-8') as f: |
|
content = f.read() |
|
|
|
# 使用字符串方法直接查找和替换,避免不必要的循环 [9](@ref) |
|
lines = content.split('\n') |
|
mtllib_modified = False |
|
|
|
for i, line in enumerate(lines): |
|
stripped_line = line.strip() |
|
if not mtllib_modified and stripped_line.startswith('map_Kd '): |
|
# 更高效的分割方式 [9](@ref) |
|
parts = line.split(' ', 1) |
|
if len(parts) > 1: |
|
old_name = parts[1].strip() |
|
new_name = f"{json_model_info.order_id}_{old_name}" |
|
lines[i] = f"map_Kd {new_name}" |
|
mtllib_modified = True |
|
print(f"已更新材质库引用: {old_name} -> {new_name}") |
|
break # 找到第一个后立即退出 |
|
|
|
# 批量写入,减少I/O操作 [6](@ref) |
|
with open(local_path, 'w', encoding='utf-8') as f: |
|
f.write('\n'.join(lines)) |
|
|
|
except IOError as e: |
|
print(f"处理文件 {local_path} 时出错: {e}") |
|
except UnicodeDecodeError as e: |
|
print(f"文件编码错误 {local_path}: {e}") |
|
|
|
print(f"下载文件: {obj_key} -> {local_path}") |
|
|
|
|
|
if download_count != 3: |
|
print(f"下载文件数量不正确: {download_count}") |
|
#直接杀死进程 |
|
os._exit(1) |
|
return True |
|
|
|
import requests |
|
import json |
|
import shutil |
|
|
|
def get_api(url): |
|
try: |
|
response = requests.get(url) |
|
response.raise_for_status() # 检查请求是否成功 |
|
response = json.loads(response.text) |
|
if response.get("code") != 1000: |
|
raise Exception(f"Error fetching URL {url}: {response.get('message')}") |
|
else: |
|
return response |
|
except requests.exceptions.RequestException as e: |
|
raise Exception(f"Error fetching URL {url}: {e}") |
|
|
|
from dataclasses import dataclass |
|
@dataclass |
|
class JSONModelInfo: |
|
obj_name: str |
|
order_id: str |
|
pid: str |
|
model_height: str |
|
print_id: str |
|
|
|
def read_pids_from_json(pid_file): |
|
"""从文件读取所有PID""" |
|
|
|
json_path = pid_file |
|
|
|
""" |
|
加载JSON文件,读取所有模型信息,应用变换后返回模型列表 |
|
""" |
|
# 检查JSON文件是否存在 |
|
if not os.path.exists(json_path): |
|
print(f"错误: JSON文件不存在 - {json_path}") |
|
return [], {}, '' |
|
|
|
# 检查是否是文件而不是目录 |
|
if not os.path.isfile(json_path): |
|
print(f"错误: 路径是目录而不是文件 - {json_path}") |
|
return [], {}, '' |
|
|
|
# 读取JSON文件 |
|
try: |
|
with open(json_path, 'r', encoding='utf-8') as f: |
|
data = json.load(f) |
|
except Exception as e: |
|
print(f"读取JSON文件失败: {e}") |
|
return [], {}, '' |
|
|
|
list_model_info = [] |
|
summary = data.get('summary') |
|
print("summary=", summary) |
|
# 安全获取 selected_machine |
|
if isinstance(summary, dict): |
|
selected_machine = summary.get('selected_machine', '') |
|
else: |
|
# 如果 summary 不是字典,尝试从根级别获取 |
|
selected_machine = data.get('selected_machine', '') |
|
print("selected_machine=", selected_machine) |
|
# 处理每个模型 |
|
for model in data.get('models', []): |
|
obj_name = model.get('file_name', '') |
|
|
|
parts = obj_name.split('_') |
|
|
|
order_id = parts[0] |
|
print_id = parts[2].replace("P", "") |
|
pid = parts[1] |
|
model_height = parts[3] |
|
|
|
model_info = JSONModelInfo( |
|
obj_name=obj_name, |
|
order_id=order_id, |
|
pid=pid, |
|
model_height=model_height, |
|
print_id=print_id |
|
) |
|
list_model_info.append(model_info) |
|
|
|
return list_model_info, data, selected_machine |
|
|
|
def download_data_by_json(model_info, workdir, oss_client ): |
|
try: |
|
pid = model_info.pid |
|
model_height = model_info.model_height |
|
# target_dir = f"{workdir}/{pid}_image" |
|
target_dir = f"{workdir}" |
|
|
|
url = f"https://mp.api.suwa3d.com/api/order/getOssSuffixByOrderId?order_id={model_info.order_id}" |
|
res = requests.get(url) |
|
|
|
data = res.json()["data"] |
|
# print("datas=",data) |
|
data = data.replace("/init_obj", "") |
|
|
|
print("target_dir=", target_dir) |
|
download_textures = DataTransfer(target_dir, f"objs/download/print/{pid}/{data}/{model_height}/", oss_client) |
|
|
|
download_textures.download_data_rename_json(model_info) |
|
# 下载后检查目标文件夹是否为空 |
|
if os.path.exists(target_dir) and not os.listdir(target_dir): |
|
shutil.rmtree(target_dir) |
|
print(f"下载后检查发现目标文件夹为空,已删除: {target_dir}") |
|
except Exception as e: |
|
print(f"卡通图片下载失败: {pid}, 错误: {str(e)}") |
|
pass |
|
|
|
def get_oss_client(cfg_path): |
|
# with open(os.path.expanduser(cfg_path), "r") as config: |
|
# cfg = yaml.safe_load(config) |
|
|
|
AccessKeyId_down = 'LTAI5tSReWm8hz7dSYxxth8f' |
|
AccessKeySecret_down = '8ywTDF9upPAtvgXtLKALY2iMYHIxdS' |
|
Endpoint_down = 'oss-cn-shanghai.aliyuncs.com' |
|
Bucket_down = 'suwa3d-securedata' |
|
|
|
oss_client = oss2.Bucket( |
|
oss2.Auth(AccessKeyId_down, AccessKeySecret_down), Endpoint_down, Bucket_down |
|
) |
|
|
|
return oss_client |
|
|
|
def download_datas_by_json(pid_file, workdir, oss_config): |
|
oss_client = get_oss_client(oss_config) |
|
|
|
# json_path = os.path.join(workdir, "3DPrintLayout.json") |
|
json_path = pid_file |
|
|
|
# 读取所有PID |
|
list_model_info, data, selected_machine = read_pids_from_json(json_path) |
|
print(f"从文件读取了 {len(list_model_info)} 个PID") |
|
need_transform = True |
|
if selected_machine=="大机型" : |
|
need_transform = False |
|
# 批量下载 |
|
for model_info in list_model_info: |
|
print(f"开始下载PID: {model_info}") |
|
download_data_by_json(model_info, workdir, oss_client,) |
|
|
|
return data, need_transform |
|
|
|
def download_transform_save_by_json(pid_file, workdir, oss_config): |
|
layout_data, need_transform = download_datas_by_json(pid_file, workdir, oss_config) |
|
|
|
original_obj_pid_dir = workdir |
|
cache_type_setting_dir = os.path.join(workdir, "arrange") |
|
Path(cache_type_setting_dir).mkdir(exist_ok=True) |
|
|
|
print(f"original_obj_pid_dir={original_obj_pid_dir}, cache_type_setting_dir={cache_type_setting_dir}") |
|
|
|
if need_transform: |
|
transform_save(layout_data, original_obj_pid_dir, cache_type_setting_dir) |
|
|
|
def transform_save(layout_data, original_obj_pid_dir, cache_type_setting_dir): |
|
meshes = [] |
|
# 小打印机380*345,需要偏移-380,-345 |
|
need_offset = True |
|
for model in layout_data["models"]: |
|
transform = model.get('transform', {}) |
|
|
|
homo_matrix = transform["homo_matrix"] # 获取存储的列表 |
|
reconstructed_matrix = np.array(homo_matrix, dtype=np.float64) |
|
|
|
obj_name = model.get('file_name', '') |
|
obj_path = os.path.join(original_obj_pid_dir, obj_name) |
|
# 加载网格 |
|
try: |
|
mesh = o3d.io.read_triangle_mesh(obj_path, enable_post_processing=True) |
|
if not mesh.has_vertices(): |
|
print(f"警告: 网格无有效顶点 - {obj_path}") |
|
continue |
|
except Exception as e: |
|
print(f"加载模型失败: {obj_path} - {e}") |
|
continue |
|
|
|
original_vertices = np.asarray(mesh.vertices) |
|
|
|
transformed_vertices = custom_mesh_transform(original_vertices, reconstructed_matrix) |
|
# 如果 need_offset 为 True,应用额外的偏移 |
|
if need_offset: |
|
# 应用偏移 (-380, -345, 0) |
|
offset = np.array([-380, -345, 0]) |
|
transformed_vertices += offset |
|
print(f"已对模型 {obj_name} 应用偏移: {offset}") |
|
|
|
mesh.vertices = o3d.utility.Vector3dVector(transformed_vertices) |
|
|
|
meshes.append(mesh) |
|
|
|
# obj_path_arrange = os.path.join(original_obj_pid_dir, "arrange") |
|
obj_path_arrange = cache_type_setting_dir |
|
if not os.path.exists(obj_path_arrange): |
|
os.mkdir(obj_path_arrange) |
|
obj_path_arrange_obj = os.path.join(obj_path_arrange, obj_name) |
|
print("obj_path_arrange_obj", obj_path_arrange_obj) |
|
mesh.compute_vertex_normals() |
|
o3d.io.write_triangle_mesh(obj_path_arrange_obj, mesh,write_triangle_uvs=True) |
|
|
|
if __name__ == "__main__": |
|
parser = argparse.ArgumentParser() |
|
|
|
parser.add_argument("--pid_file", type=str, required=True, help="批次号, 也是json文件名") |
|
parser.add_argument("--workdir", type=str, required=True, help="本代码文件所在的目录") |
|
parser.add_argument("--oss_config", type=str, default="") |
|
args = parser.parse_args() |
|
|
|
download_transform_save_by_json(args.pid_file, args.workdir, args.oss_config) |
|
|
|
|