目录

DUSt3R 三维重建

报错RecursionError: maximum recursion depth exceeded in comparison

报错 numpy.core.multiarray failed to import

报错Numpy is not available 解决

升级版mast3r 速度变慢

修改了参数设置脚本：

测试效果

操作技巧

---

DUSt3R 三维重建

git地址：

GitHub - naver/dust3r: DUSt3R: Geometric 3D Vision Made Easy

报错RecursionError: maximum recursion depth exceeded in comparison

解决方法：

原版给的python版本是3.11，3.10换成3.11问题消失了。

Warning, cannot find cuda-compiled version of RoPE2D, using a slow pytorch version instead

报错 numpy.core.multiarray failed to import

Traceback (most recent call last):
  File "F:\project\3dgs\dust3r\aaa.py", line 1, in <module>
    import cv2
  File "D:\ProgramData\miniconda3\envs\py310\lib\site-packages\cv2\__init__.py", line 181, in <module>
    bootstrap()
  File "D:\ProgramData\miniconda3\envs\py310\lib\site-packages\cv2\__init__.py", line 153, in bootstrap
    native_module = importlib.import_module("cv2")
  File "D:\ProgramData\miniconda3\envs\py310\lib\importlib\__init__.py", line 126, in import_module
    return _bootstrap._gcd_import(name[level:], package, level)
ImportError: numpy.core.multiarray failed to import

pip uninstall numpy opencv-python
pip install numpy opencv-python

报错Numpy is not available 解决

File "F:\project\3dgs\dust3r\aaa.py", line 6, in <module> torch.from_numpy(np.array([1,3,5],np.uint8))

RuntimeError: Numpy is not available

解决方法：

pip uninstall numpy

pip install numpy==1.26.0

解决RuntimeError: Numpy is not available-CSDN博客

升级版mast3r 速度变慢

修改了参数设置脚本：

def get_args_parser():
    parser = argparse.ArgumentParser()
    parser_url = parser.add_mutually_exclusive_group()
    parser_url.add_argument("--local_network", action='store_true', default='0.0.0.0',
                            help="make app accessible on local network: address will be set to 0.0.0.0")
    parser_url.add_argument("--server_name", type=str, default='127.0.0.1', help="server url, default is 127.0.0.1")
    parser.add_argument("--image_size", type=int, default=512, choices=[512, 224], help="image size")
    parser.add_argument("--server_port", type=int,default= 7860, help="7860.")
    # parser_weights = parser.add_mutually_exclusive_group(required=True)
    # parser_weights.add_argument("--weights", type=str, help="path to the model weights", default=r'F:\project\3dgs\ViewCrafter-main\checkpoints\DUSt3R_ViTLarge_BaseDecoder_512_dpt.pth')
    # parser_weights.add_argument("--model_name", type=str, help="name of the model weights",default="DUSt3R_ViTLarge_BaseDecoder_512_dpt")
                                # choices=["DUSt3R_ViTLarge_BaseDecoder_512_dpt",
                                #          "DUSt3R_ViTLarge_BaseDecoder_512_linear",
                                #          "DUSt3R_ViTLarge_BaseDecoder_224_linear"])
    parser.add_argument("--weights", type=str, help="path to the model weights", default=r'F:\project\3dgs\ViewCrafter-main\checkpoints\DUSt3R_ViTLarge_BaseDecoder_512_dpt.pth')
    parser.add_argument("--model_name", type=str, help="name of the model weights",default="DUSt3R_ViTLarge_BaseDecoder_512_dpt")
    parser.add_argument("--device", type=str, default='cuda', help="pytorch device")
    parser.add_argument("--tmp_dir", type=str, default=None, help="value for tempfile.tempdir")
    parser.add_argument("--silent", action='store_true', default=False,
                        help="silence logs")
    return parser

测试效果

Our new model with a local feature head, metric pointmaps, and a more scalable global alignment!

我们的新模型具有局部特征头、度量点图和更具可扩展性的全局对齐功能

模型2.6G

找了两个图片测试，速度比DUSt3R慢了。

在驾驶场景效果也没看到提升.

点云：

操作技巧

浏览器操作，滚轮可以缩放点云的大小，不能缩放可能要换个浏览器。