Multimodal Industrial Anomaly Detection by Crossmodal Feature Mapping (CVPR 2024)在TITAN上曲折复现-笔记

最新推荐文章于 2026-08-20 17:57:47 发布
原创 最新推荐文章于 2026-08-20 17:57:47 发布 · 1.2k 阅读 · 13 ·
本内容遵循CC 4.0 BY-SA版权协议
版权声明：本文为博主原创文章，遵循 CC 4.0 BY-SA 版权协议，转载请附上原文出处链接和本声明。
GEO检测

文章目录

参考文献

wandb快速上手、使用心得（超好用的Tensorboard高替品）

按给出的 requirements.txt建立（失败）

创建环境conda create -n zgp_crosspy311 python=3.10

网络（能连网省略）

(zgp_crosspy311) ~ cd /home/cszx/PySocks
python setup.py install
python D:\ssh-reverse-socket-master\socket5_proxy.py --ssh_cmd “cszx@172.29.6.20 -p 22” --socket5_port 8080

安装wandb

pip install wandb -i https://pypi.tuna.tsinghua.edu.cn/simple

注册

wandb官网
在这里插入图片描述

登录wandb login
wandb: Logging into wandb.ai. (Learn how to deploy a W&B server locally: https://wandb.me/wandb-server)
wandb: You can find your API key in your browser here: https://wandb.ai/authorize
wandb: Paste an API key from your profile and hit enter, or press ctrl+c to quit: 
wandb: Appending key for api.wandb.ai to your netrc file: /home/cszx/.netrc

可以了

pip install -r requirements.txt

注释掉重复的
在这里插入图片描述

Could not find a version that satisfies the requirement open3d==0.18.0 (from versions: none)
取消版本指定，放到最后
报错

ERROR: Ignored the following versions that require a different python version: 1.6.2 Requires-Python >=3.7,<3.10; 1.6.3 Requires-Python >=3.7,<3.10; 1.7.0 Requires-Python >=3.7,<3.10; 1.7.1 Requires-Python >=3.7,<3.10
ERROR: Could not find a version that satisfies the requirement open3d (from versions: none)
ERROR: No matching distribution found for open3d

删除环境，重建py3.9（误判，没什么用）

conda deactivate
conda remove --name zgp_crosspy311 --all
conda create -n zgp_crosspy39 python=3.9

再次pip install -r requirements.txt，又报错需要Python >=3.10

ERROR: Ignored the following versions that require a different python version: 1.14.0 Requires-Python >=3.10; 1.14.0rc1 Requires-Python >=3.10; 1.14.0rc2 Requires-Python >=3.10; 1.14.1 Requires-Python >=3.10; 2.1.0 Requires-Python >=3.10; 2.1.0rc1 Requires-Python >=3.10; 2.1.1 Requires-Python >=3.10
ERROR: Could not find a version that satisfies the requirement open3d==0.18.0 (from versions: none)
ERROR: No matching distribution found for open3d==0.18.0

醉了，取消对 open3d版本指定，再次pip install -r requirements.txt
还是一样
发现open3d只在预处理代码中使用，放弃安装open3d

Installing collected packages: scikit_learn, pandas, matplotlib, huggingface_hub, GitPython, wandb, torchvision, timm
Successfully installed GitPython-3.1.43 huggingface_hub-0.25.0 matplotlib-3.8.3 pandas-1.5.3 scikit_learn-1.2.1 timm-1.0.3 torchvision-0.14.1 wandb-0.15.5

pip install “git+git://github.com/erikwijmans/Pointnet2_PyTorch.git#egg=pointnet2_ops&subdirectory=pointnet2_ops_lib”

报错

× git clone --quiet git://github.com/erikwijmans/Pointnet2_PyTorch.git /tmp/pip-install-9tay989u/pointnet2-ops_f7333ff17ba744cf9583c9079b60f683 did not run successfully.
│ exit code: 128
╰─> See above for output.
改为pip install “git+https://github.com/erikwijmans/Pointnet2_PyTorch.git#egg=pointnet2_ops&subdirectory=pointnet2_ops_lib”

报错

  × python setup.py egg_info did not run successfully.
  │ exit code: 1
  ╰─> [10 lines of output]
      Traceback (most recent call last):
        File "<string>", line 2, in <module>
        File "<pip-setuptools-caller>", line 34, in <module>
        File "/tmp/pip-install-gexgid8e/pointnet2-ops_880964657f81407eb85e9d7551af7160/pointnet2_ops_lib/setup.py", line 6, in <module>
          from torch.utils.cpp_extension import BuildExtension, CUDAExtension
        File "/home/cszx/miniconda3/envs/zgp_crosspy39/lib/python3.9/site-packages/torch/utils/__init__.py", line 4, in <module>
          from .throughput_benchmark import ThroughputBenchmark
        File "/home/cszx/miniconda3/envs/zgp_crosspy39/lib/python3.9/site-packages/torch/utils/throughput_benchmark.py", line 2, in <module>
          import torch._C
      ModuleNotFoundError: No module named 'torch._C'
      [end of output]
      ```
离线安装python3 -m pip install /home/cszx/c1/zgp/crossfm/Pointnet2_PyTorch-master/Pointnet2_PyTorch-master/pointnet2_ops_lib
也一样

直接运行
报错

Traceback (most recent call last):
  File "/home/cszx/c1/zgp/crossfm/cfm_inference.py", line 4, in <module>
    from torchvision import transforms
  File "/home/cszx/miniconda3/envs/zgp_crosspy39/lib/python3.9/site-packages/torchvision/__init__.py", line 5, in <module>
    from torchvision import datasets, io, models, ops, transforms, utils
  File "/home/cszx/miniconda3/envs/zgp_crosspy39/lib/python3.9/site-packages/torchvision/datasets/__init__.py", line 1, in <module>
    from ._optical_flow import FlyingChairs, FlyingThings3D, HD1K, KittiFlow, Sintel
  File "/home/cszx/miniconda3/envs/zgp_crosspy39/lib/python3.9/site-packages/torchvision/datasets/_optical_flow.py", line 11, in <module>
    from ..io.image import _read_png_16
  File "/home/cszx/miniconda3/envs/zgp_crosspy39/lib/python3.9/site-packages/torchvision/io/__init__.py", line 5, in <module>
    from ..utils import _log_api_usage_once
  File "/home/cszx/miniconda3/envs/zgp_crosspy39/lib/python3.9/site-packages/torchvision/utils.py", line 23, in <module>
    @torch.no_grad()
AttributeError: module 'torch' has no attribute 'no_grad'

不太行

鉴于作者致谢了M3DM，放弃安装其 requirements.txt环境，改为使用M3DM环境

conda create -n zgp_cross_fm3dm38 --clone zgp_m3dm

安装wandb

pip install wandb -i https://pypi.tuna.tsinghua.edu.cn/simple

wandb login --relogin

后续会提示链接地址

设置远程编译路径

在这里插入图片描述

如果无法设置，先从Develop处设置好连接

cfm_inference.py 测试，生成anomaly maps

设置参数
–dataset_path:数据集根目录/home/cszx/c1/zgp/mvtec3d
–checkpoint_folder: 原checkpoints/checkpoints_CFM_mvtec.
–class_name:原
–epochs_no: 50改为5
–batch_size: 4
–qualitative_folder: 建立results/qualitatives_mvtec文件夹
–quantitative_folder: 建立results/quantitatives_mvtec文件夹
–visualize_plot: False
–produce_qualitatives: False
Checkpoints放好
在这里插入图片描述
修改02_eval_mvtec.sh
250轮改为2

sh 02_eval_mvtec.sh
报错

sh 02_eval_mvtec.sh
Traceback (most recent call last):
  File "/home/cszx/miniconda3/envs/zgp_m3dm/lib/python3.8/site-packages/PySocks-1.7.1-py3.8.egg/socks.py", line 809, in connect
  File "/home/cszx/miniconda3/envs/zgp_m3dm/lib/python3.8/site-packages/PySocks-1.7.1-py3.8.egg/socks.py", line 443, in _negotiate_SOCKS5
  File "/home/cszx/miniconda3/envs/zgp_m3dm/lib/python3.8/site-packages/PySocks-1.7.1-py3.8.egg/socks.py", line 524, in _SOCKS5_request
  File "/home/cszx/miniconda3/envs/zgp_m3dm/lib/python3.8/site-packages/PySocks-1.7.1-py3.8.egg/socks.py", line 276, in _readall
  File "/home/cszx/miniconda3/envs/zgp_m3dm/lib/python3.8/socket.py", line 669, in readinto
    return self._sock.recv_into(b)
socket.timeout: timed out

直接运行
报错

Traceback (most recent call last):
  File "/home/cszx/c1/zgp/crossfm/cfm_inference.py", line 225, in <module>
    infer_CFM(args)
  File "/home/cszx/c1/zgp/crossfm/cfm_inference.py", line 33, in infer_CFM
    test_loader = get_data_loader("test", class_name = args.class_name, img_size = 224, dataset_path = args.dataset_path)
  File "/home/cszx/c1/zgp/crossfm/models/dataset.py", line 173, in get_data_loader
    dataset = TestDataset(class_name = class_name, img_size = img_size, dataset_path = dataset_path)
  File "/home/cszx/c1/zgp/crossfm/models/dataset.py", line 98, in __init__
    super().__init__(split = "test", class_name = class_name, img_size = img_size, dataset_path = dataset_path)
  File "/home/cszx/c1/zgp/crossfm/models/dataset.py", line 48, in __init__
    self.img_path = os.path.join(dataset_path, self.cls, split)
  File "/home/cszx/miniconda3/envs/zgp_cross_fm3dm38/lib/python3.8/posixpath.py", line 90, in join
    genericpath._check_arg_types('join', a, *p)
  File "/home/cszx/miniconda3/envs/zgp_cross_fm3dm38/lib/python3.8/genericpath.py", line 152, in _check_arg_types
    raise TypeError(f'{funcname}() argument must be str, bytes, or '
TypeError: join() argument must be str, bytes, or os.PathLike object, not 'NoneType'

ssh://cszx@172.29.6.20:22/home/cszx/miniconda3/envs/zgp_cross_fm3dm38/bin/python3.8 -u /home/cszx/c1/zgp/crossfm/cfm_inference.py
Traceback (most recent call last):
  File "/home/cszx/c1/zgp/crossfm/cfm_inference.py", line 225, in <module>
    infer_CFM(args)
  File "/home/cszx/c1/zgp/crossfm/cfm_inference.py", line 33, in infer_CFM
    test_loader = get_data_loader("test", class_name = args.class_name, img_size = 224, dataset_path = args.dataset_path)
  File "/home/cszx/c1/zgp/crossfm/models/dataset.py", line 173, in get_data_loader
    dataset = TestDataset(class_name = class_name, img_size = img_size, dataset_path = dataset_path)
  File "/home/cszx/c1/zgp/crossfm/models/dataset.py", line 98, in __init__
    super().__init__(split = "test", class_name = class_name, img_size = img_size, dataset_path = dataset_path)
  File "/home/cszx/c1/zgp/crossfm/models/dataset.py", line 48, in __init__
    self.img_path = os.path.join(dataset_path, self.cls, split)
  File "/home/cszx/miniconda3/envs/zgp_cross_fm3dm38/lib/python3.8/posixpath.py", line 90, in join
    genericpath._check_arg_types('join', a, *p)
  File "/home/cszx/miniconda3/envs/zgp_cross_fm3dm38/lib/python3.8/genericpath.py", line 152, in _check_arg_types
    raise TypeError(f'{funcname}() argument must be str, bytes, or '
TypeError: join() argument must be str, bytes, or os.PathLike object, not 'NoneType'

参数问题
数据集路径改正
将class_name由None改为parser.add_argument(‘–class_name’, default = “bagel”,
报错

ssh://cszx@172.29.6.20:22/home/cszx/miniconda3/envs/zgp_cross_fm3dm38/bin/python3.8 -u /home/cszx/c1/zgp/crossfm/cfm_inference.py
Traceback (most recent call last):
  File "/home/cszx/miniconda3/envs/zgp_cross_fm3dm38/lib/python3.8/site-packages/urllib3/connection.py", line 198, in _new_conn
    sock = connection.create_connection(
  File "/home/cszx/miniconda3/envs/zgp_cross_fm3dm38/lib/python3.8/site-packages/urllib3/util/connection.py", line 60, in create_connection
    for res in socket.getaddrinfo(host, port, family, socket.SOCK_STREAM):
  File "/home/cszx/miniconda3/envs/zgp_cross_fm3dm38/lib/python3.8/socket.py", line 918, in getaddrinfo
    for res in _socket.getaddrinfo(host, port, family, type, proto, flags):
socket.gaierror: [Errno -2] Name or service not known

The above exception was the direct cause of the following exception:

Traceback (most recent call last):
  File "/home/cszx/miniconda3/envs/zgp_cross_fm3dm38/lib/python3.8/site-packages/urllib3/connectionpool.py", line 793, in urlopen
    response = self._make_request(
  File "/home/cszx/miniconda3/envs/zgp_cross_fm3dm38/lib/python3.8/site-packages/urllib3/connectionpool.py", line 491, in _make_request
    raise new_e
  File "/home/cszx/miniconda3/envs/zgp_cross_fm3dm38/lib/python3.8/site-packages/urllib3/connectionpool.py", line 467, in _make_request
    self._validate_conn(conn)
  File "/home/cszx/miniconda3/envs/zgp_cross_fm3dm38/lib/python3.8/site-packages/urllib3/connectionpool.py", line 1099, in _validate_conn
    conn.connect()
  File "/home/cszx/miniconda3/envs/zgp_cross_fm3dm38/lib/python3.8/site-packages/urllib3/connection.py", line 616, in connect
    self.sock = sock = self._new_conn()
  File "/home/cszx/miniconda3/envs/zgp_cross_fm3dm38/lib/python3.8/site-packages/urllib3/connection.py", line 205, in _new_conn
    raise NameResolutionError(self.host, self, e) from e
urllib3.exceptions.NameResolutionError: <urllib3.connection.HTTPSConnection object at 0x7f0eeeb4cc10>: Failed to resolve 'huggingface.co' ([Errno -2] Name or service not known)

The above exception was the direct cause of the following exception:

Traceback (most recent call last):
  File "/home/cszx/miniconda3/envs/zgp_cross_fm3dm38/lib/python3.8/site-packages/requests/adapters.py", line 589, in send
    resp = conn.urlopen(
  File "/home/cszx/miniconda3/envs/zgp_cross_fm3dm38/lib/python3.8/site-packages/urllib3/connectionpool.py", line 847, in urlopen
    retries = retries.increment(
  File "/home/cszx/miniconda3/envs/zgp_cross_fm3dm38/lib/python3.8/site-packages/urllib3/util/retry.py", line 515, in increment
    raise MaxRetryError(_pool, url, reason) from reason  # type: ignore[arg-type]
urllib3.exceptions.MaxRetryError: HTTPSConnectionPool(host='huggingface.co', port=443): Max retries exceeded with url: /timm/vit_base_patch8_224.dino/resolve/main/pytorch_model.bin (Caused by NameResolutionError("<urllib3.connection.HTTPSConnection object at 0x7f0eeeb4cc10>: Failed to resolve 'huggingface.co' ([Errno -2] Name or service not known)"))

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/home/cszx/miniconda3/envs/zgp_cross_fm3dm38/lib/python3.8/site-packages/huggingface_hub/file_download.py", line 1722, in _get_metadata_or_catch_error
    metadata = get_hf_file_metadata(url=url, proxies=proxies, timeout=etag_timeout, headers=headers)
  File "/home/cszx/miniconda3/envs/zgp_cross_fm3dm38/lib/python3.8/site-packages/huggingface_hub/utils/_validators.py", line 114, in _inner_fn
    return fn(*args, **kwargs)
  File "/home/cszx/miniconda3/envs/zgp_cross_fm3dm38/lib/python3.8/site-packages/huggingface_hub/file_download.py", line 1645, in get_hf_file_metadata
    r = _request_wrapper(
  File "/home/cszx/miniconda3/envs/zgp_cross_fm3dm38/lib/python3.8/site-packages/huggingface_hub/file_download.py", line 372, in _request_wrapper
    response = _request_wrapper(
  File "/home/cszx/miniconda3/envs/zgp_cross_fm3dm38/lib/python3.8/site-packages/huggingface_hub/file_download.py", line 395, in _request_wrapper
    response = get_session().request(method=method, url=url, **params)
  File "/home/cszx/miniconda3/envs/zgp_cross_fm3dm38/lib/python3.8/site-packages/requests/sessions.py", line 589, in request
    resp = self.send(prep, **send_kwargs)
  File "/home/cszx/miniconda3/envs/zgp_cross_fm3dm38/lib/python3.8/site-packages/requests/sessions.py", line 703, in send
    r = adapter.send(request, **kwargs)
  File "/home/cszx/miniconda3/envs/zgp_cross_fm3dm38/lib/python3.8/site-packages/huggingface_hub/utils/_http.py", line 66, in send
    return super().send(request, *args, **kwargs)
  File "/home/cszx/miniconda3/envs/zgp_cross_fm3dm38/lib/python3.8/site-packages/requests/adapters.py", line 622, in send
    raise ConnectionError(e, request=request)
requests.exceptions.ConnectionError: (MaxRetryError('HTTPSConnectionPool(host=\'huggingface.co\', port=443): Max retries exceeded with url: /timm/vit_base_patch8_224.dino/resolve/main/pytorch_model.bin (Caused by NameResolutionError("<urllib3.connection.HTTPSConnection object at 0x7f0eeeb4cc10>: Failed to resolve \'huggingface.co\' ([Errno -2] Name or service not known)"))'), '(Request ID: 41b259f8-838f-467f-88c7-74491b49fe8a)')

The above exception was the direct cause of the following exception:

Traceback (most recent call last):
  File "/home/cszx/c1/zgp/crossfm/cfm_inference.py", line 226, in <module>
    infer_CFM(args)
  File "/home/cszx/c1/zgp/crossfm/cfm_inference.py", line 37, in infer_CFM
    feature_extractor = MultimodalFeatures()
  File "/home/cszx/c1/zgp/crossfm/models/features.py", line 20, in __init__
    self.deep_feature_extractor = FeatureExtractors(device = self.device, 
  File "/home/cszx/c1/zgp/crossfm/models/full_models.py", line 24, in __init__
    self.rgb_backbone = timm.create_model(model_name = rgb_backbone_name, pretrained = True, **kwargs)
  File "/home/cszx/miniconda3/envs/zgp_cross_fm3dm38/lib/python3.8/site-packages/timm/models/_factory.py", line 117, in create_model
    model = create_fn(
  File "/home/cszx/miniconda3/envs/zgp_cross_fm3dm38/lib/python3.8/site-packages/timm/models/vision_transformer.py", line 2066, in vit_base_patch8_224
    model = _create_vision_transformer('vit_base_patch8_224', pretrained=pretrained, **dict(model_args, **kwargs))
  File "/home/cszx/miniconda3/envs/zgp_cross_fm3dm38/lib/python3.8/site-packages/timm/models/vision_transformer.py", line 1946, in _create_vision_transformer
    return build_model_with_cfg(
  File "/home/cszx/miniconda3/envs/zgp_cross_fm3dm38/lib/python3.8/site-packages/timm/models/_builder.py", line 418, in build_model_with_cfg
    load_pretrained(
  File "/home/cszx/miniconda3/envs/zgp_cross_fm3dm38/lib/python3.8/site-packages/timm/models/_builder.py", line 196, in load_pretrained
    state_dict = load_state_dict_from_hf(pretrained_loc)
  File "/home/cszx/miniconda3/envs/zgp_cross_fm3dm38/lib/python3.8/site-packages/timm/models/_hub.py", line 188, in load_state_dict_from_hf
    cached_file = hf_hub_download(hf_model_id, filename=filename, revision=hf_revision)
  File "/home/cszx/miniconda3/envs/zgp_cross_fm3dm38/lib/python3.8/site-packages/huggingface_hub/utils/_validators.py", line 114, in _inner_fn
    return fn(*args, **kwargs)
  File "/home/cszx/miniconda3/envs/zgp_cross_fm3dm38/lib/python3.8/site-packages/huggingface_hub/file_download.py", line 1221, in hf_hub_download
    return _hf_hub_download_to_cache_dir(
  File "/home/cszx/miniconda3/envs/zgp_cross_fm3dm38/lib/python3.8/site-packages/huggingface_hub/file_download.py", line 1325, in _hf_hub_download_to_cache_dir
    _raise_on_head_call_error(head_call_error, force_download, local_files_only)
  File "/home/cszx/miniconda3/envs/zgp_cross_fm3dm38/lib/python3.8/site-packages/huggingface_hub/file_download.py", line 1826, in _raise_on_head_call_error
    raise LocalEntryNotFoundError(
huggingface_hub.utils._errors.LocalEntryNotFoundError: An error happened while trying to locate the file on the Hub and we cannot find the requested files in the local cache. Please check your connection and try again or make sure your Internet connection is on.

Process finished with exit code 1
设置网络后仍找不到
 raise LocalEntryNotFoundError(
huggingface_hub.utils._errors.LocalEntryNotFoundError: An error happened while trying to locate the file on the Hub and we cannot find the requested files in the local cache. Please check your connection and try again or make sure your Internet connection is on.

改为使用下载的

    #self.rgb_backbone = timm.create_model(model_name = rgb_backbone_name, pretrained = True, **kwargs)
    self.rgb_backbone = timm.create_model(model_name=rgb_backbone_name, pretrained=True, pretrained_cfg_overlay=dict( file='dino_vitbase8_pretrain.pth'),
                                       **kwargs)

No such file or directory: ‘./checkpoints/checkpoints_CFM_mvtec/bagel/CFM_2Dto3D_bagel_5ep_4bs.pth’

改为
parser.add_argument(‘–epochs_no’, default = 50, type = int,
help = ‘Number of epochs to train the CFMs.’)
成功了

ssh://cszx@172.29.6.20:22/home/cszx/miniconda3/envs/zgp_cross_fm3dm38/bin/python3.8 -u /home/cszx/c1/zgp/crossfm/cfm_inference.py
Extracting feature from class: bagel.: 100%|██| 110/110 [00:19<00:00,  5.68it/s]
Metrics for class bagel with 50ep_4bs
AUPRO@30% | AUPRO@10% | AUPRO@5% | AUPRO@1% | P-AUROC | I-AUROC
  0.980   |   0.942   |   0.885  |   0.473  |   0.997 |   0.988

Process finished with exit code 0

三次相同

批量运行

No such file or directory: ‘./checkpoints/checkpoints_CFM_mvtec/carrot/CFM_2Dto3D_carrot_2ep_4bs.pth’
设置epochs=50
batch_size=4

 sh 02_eval_mvtec.sh
Extracting feature from class: bagel.: 100%|███████████████████████████████████████████████████████████████████████████████████████████| 110/110 [00:19<00:00,  5.72it/s]
Metrics for class bagel with 50ep_4bs
AUPRO@30% | AUPRO@10% | AUPRO@5% | AUPRO@1% | P-AUROC | I-AUROC
  0.980   |   0.942   |   0.885  |   0.473  |   0.997 |   0.988
Extracting feature from class: cable_gland.: 100%|█████████████████████████████████████████████████████████████████████████████████████| 108/108 [00:15<00:00,  7.09it/s]
Metrics for class cable_gland with 50ep_4bs
AUPRO@30% | AUPRO@10% | AUPRO@5% | AUPRO@1% | P-AUROC | I-AUROC
  0.954   |   0.866   |   0.765  |   0.368  |   0.986 |   0.871
Extracting feature from class: carrot.: 100%|██████████████████████████████████████████████████████████████████████████████████████████| 159/159 [00:20<00:00,  7.67it/s]
Metrics for class carrot with 50ep_4bs
AUPRO@30% | AUPRO@10% | AUPRO@5% | AUPRO@1% | P-AUROC | I-AUROC
  0.982   |   0.946   |   0.893  |   0.486  |   0.998 |   0.942
Extracting feature from class: cookie.: 100%|██████████████████████████████████████████████████████████████████████████████████████████| 131/131 [00:22<00:00,  5.73it/s]
Metrics for class cookie with 50ep_4bs
AUPRO@30% | AUPRO@10% | AUPRO@5% | AUPRO@1% | P-AUROC | I-AUROC
  0.946   |   0.901   |   0.845  |   0.477  |   0.972 |   0.990
Extracting feature from class: dowel.: 100%|███████████████████████████████████████████████████████████████████████████████████████████| 130/130 [00:16<00:00,  7.66it/s]
Metrics for class dowel with 50ep_4bs
AUPRO@30% | AUPRO@10% | AUPRO@5% | AUPRO@1% | P-AUROC | I-AUROC
  0.930   |   0.800   |   0.699  |   0.347  |   0.978 |   0.815
Extracting feature from class: foam.: 100%|████████████████████████████████████████████████████████████████████████████████████████████| 100/100 [00:17<00:00,  5.82it/s]
Metrics for class foam with 50ep_4bs
AUPRO@30% | AUPRO@10% | AUPRO@5% | AUPRO@1% | P-AUROC | I-AUROC
  0.937   |   0.830   |   0.723  |   0.342  |   0.984 |   0.795
Extracting feature from class: peach.: 100%|███████████████████████████████████████████████████████████████████████████████████████████| 132/132 [00:19<00:00,  6.84it/s]
Metrics for class peach with 50ep_4bs
AUPRO@30% | AUPRO@10% | AUPRO@5% | AUPRO@1% | P-AUROC | I-AUROC
  0.980   |   0.940   |   0.880  |   0.461  |   0.998 |   0.939
Extracting feature from class: potato.: 100%|██████████████████████████████████████████████████████████████████████████████████████████| 114/114 [00:15<00:00,  7.47it/s]
Metrics for class potato with 50ep_4bs
AUPRO@30% | AUPRO@10% | AUPRO@5% | AUPRO@1% | P-AUROC | I-AUROC
  0.982   |   0.946   |   0.892  |   0.480  |   0.997 |   0.896
Extracting feature from class: rope.: 100%|████████████████████████████████████████████████████████████████████████████████████████████| 101/101 [00:13<00:00,  7.57it/s]
Metrics for class rope with 50ep_4bs
AUPRO@30% | AUPRO@10% | AUPRO@5% | AUPRO@1% | P-AUROC | I-AUROC
  0.952   |   0.859   |   0.757  |   0.356  |   0.994 |   0.591
Extracting feature from class: tire.: 100%|████████████████████████████████████████████████████████████████████████████████████████████| 112/112 [00:15<00:00,  7.24it/s]
Metrics for class tire with 50ep_4bs
AUPRO@30% | AUPRO@10% | AUPRO@5% | AUPRO@1% | P-AUROC | I-AUROC
  0.969   |   0.906   |   0.819  |   0.392  |   0.991 |   0.703

设置epochs=250
batch_size=4
python /home/cszx/c1/zgp/crossfm/cfm_inference.py

报错

No checkpoint found at 'checkpoints/dino_vitbase8_pretrain.pth'
Traceback (most recent call last):
  File "/home/cszx/c1/zgp/crossfm/cfm_inference.py", line 226, in <module>
    infer_CFM(args)
  File "/home/cszx/c1/zgp/crossfm/cfm_inference.py", line 37, in infer_CFM
    feature_extractor = MultimodalFeatures()
  File "/home/cszx/c1/zgp/crossfm/models/features.py", line 20, in __init__
    self.deep_feature_extractor = FeatureExtractors(device = self.device, 
  File "/home/cszx/c1/zgp/crossfm/models/full_models.py", line 25, in __init__
    self.rgb_backbone = timm.create_model(model_name=rgb_backbone_name, pretrained=True, pretrained_cfg_overlay=dict( file='checkpoints/dino_vitbase8_pretrain.pth'),
  File "/home/cszx/miniconda3/envs/zgp_cross_fm3dm38/lib/python3.8/site-packages/timm/models/_factory.py", line 117, in create_model
    model = create_fn(
  File "/home/cszx/miniconda3/envs/zgp_cross_fm3dm38/lib/python3.8/site-packages/timm/models/vision_transformer.py", line 2066, in vit_base_patch8_224
    model = _create_vision_transformer('vit_base_patch8_224', pretrained=pretrained, **dict(model_args, **kwargs))
  File "/home/cszx/miniconda3/envs/zgp_cross_fm3dm38/lib/python3.8/site-packages/timm/models/vision_transformer.py", line 1946, in _create_vision_transformer
    return build_model_with_cfg(
  File "/home/cszx/miniconda3/envs/zgp_cross_fm3dm38/lib/python3.8/site-packages/timm/models/_builder.py", line 418, in build_model_with_cfg
    load_pretrained(
  File "/home/cszx/miniconda3/envs/zgp_cross_fm3dm38/lib/python3.8/site-packages/timm/models/_builder.py", line 168, in load_pretrained
    state_dict = load_state_dict(pretrained_loc)
  File "/home/cszx/miniconda3/envs/zgp_cross_fm3dm38/lib/python3.8/site-packages/timm/models/_helpers.py", line 59, in load_state_dict
    raise FileNotFoundError()
FileNotFoundError

cd /home/cszx/c1/zgp/crossfm
(zgp_cross_fm3dm38) ~/c1/zgp/crossfm python /home/cszx/c1/zgp/crossfm/cfm_inference.py
Metrics for class bagel with 250ep_4bs
AUPRO@30% | AUPRO@10% | AUPRO@5% | AUPRO@1% | P-AUROC | I-AUROC
0.981 | 0.942 | 0.885 | 0.476 | 0.997 | 0.992

sh 02_eval_mvtec.sh
Extracting feature from class: bagel.: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 110/110 [00:19<00:00, 5.77it/s]
Metrics for class bagel with 250ep_4bs
AUPRO@30% | AUPRO@10% | AUPRO@5% | AUPRO@1% | P-AUROC | I-AUROC
0.981 | 0.942 | 0.885 | 0.476 | 0.997 | 0.992
Extracting feature from class: cable_gland.: 100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 108/108 [00:15<00:00, 7.19it/s]
Metrics for class cable_gland with 250ep_4bs
AUPRO@30% | AUPRO@10% | AUPRO@5% | AUPRO@1% | P-AUROC | I-AUROC
0.958 | 0.878 | 0.785 | 0.385 | 0.987 | 0.896
Extracting feature from class: carrot.: 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 159/159 [00:20<00:00, 7.74it/s]
Metrics for class carrot with 250ep_4bs
AUPRO@30% | AUPRO@10% | AUPRO@5% | AUPRO@1% | P-AUROC | I-AUROC
0.982 | 0.946 | 0.893 | 0.484 | 0.998 | 0.947
Extracting feature from class: cookie.: 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 131/131 [00:22<00:00, 5.72it/s]
Metrics for class cookie with 250ep_4bs
AUPRO@30% | AUPRO@10% | AUPRO@5% | AUPRO@1% | P-AUROC | I-AUROC
0.946 | 0.902 | 0.847 | 0.485 | 0.972 | 0.990
Extracting feature from class: dowel.: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 130/130 [00:16<00:00, 7.68it/s]
Metrics for class dowel with 250ep_4bs
AUPRO@30% | AUPRO@10% | AUPRO@5% | AUPRO@1% | P-AUROC | I-AUROC
0.935 | 0.812 | 0.713 | 0.353 | 0.980 | 0.771
Extracting feature from class: foam.: 100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 100/100 [00:17<00:00, 5.84it/s]
Metrics for class foam with 250ep_4bs
AUPRO@30% | AUPRO@10% | AUPRO@5% | AUPRO@1% | P-AUROC | I-AUROC
0.936 | 0.827 | 0.725 | 0.342 | 0.984 | 0.796
Extracting feature from class: peach.: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 132/132 [00:19<00:00, 6.82it/s]
Metrics for class peach with 250ep_4bs
AUPRO@30% | AUPRO@10% | AUPRO@5% | AUPRO@1% | P-AUROC | I-AUROC
0.980 | 0.941 | 0.883 | 0.464 | 0.998 | 0.944
Extracting feature from class: potato.: 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 114/114 [00:15<00:00, 7.48it/s]
Metrics for class potato with 250ep_4bs
AUPRO@30% | AUPRO@10% | AUPRO@5% | AUPRO@1% | P-AUROC | I-AUROC
0.982 | 0.947 | 0.894 | 0.483 | 0.998 | 0.913
Extracting feature from class: rope.: 100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 101/101 [00:13<00:00, 7.63it/s]
Metrics for class rope with 250ep_4bs
AUPRO@30% | AUPRO@10% | AUPRO@5% | AUPRO@1% | P-AUROC | I-AUROC
0.946 | 0.843 | 0.737 | 0.339 | 0.992 | 0.637
Extracting feature from class: tire.: 100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 112/112 [00:15<00:00, 7.23it/s]
Metrics for class tire with 250ep_4bs
AUPRO@30% | AUPRO@10% | AUPRO@5% | AUPRO@1% | P-AUROC | I-AUROC
0.972 | 0.916 | 0.838 | 0.415 | 0.993 | 0.783

标签
#笔记
桂萍
关注
from pointnet2_ops import pointnet2_utils 安装过程
goodgoodstudy___的博客
12-16 8335
pointnet2_ops
PointMamba复现安装Pointnet++中包报错的解决方案
kalimera的博客
11-11 1254
PointMamba复现报错及解决方案（1）没有安装git。（2）whl build失败
Multimodal Industrial Anomaly Detection by Crossmodal Feature Mapping (CVPR 2024)在3080上复现-笔记
10-25 1454
修改数据集路径和class_name parser.add_argument(‘–dataset_path’, default = ‘/home/zgp/Documents/m3dmpre/datasets/mvtec3d’, type = str,参考《远程服务器上，再次配笔记1、2、11、12相同的深度学习和gcc环境》上传修改了checkpoint地址的文件覆盖之前的文件。基于veh_sem虚拟环境安装。注释掉torch单独安装。
Python-python常用方法utils1.zip
02-24
python常用方法utils
Multimodal Industrial Anomaly Detection by Crossmodal Feature Mapping 论文精读
oj_geen的博客
04-01 1624
工业异常检测（AD）旨在识别产品中的异常特征或缺陷，是质量检测流程的关键部分。由于异常的罕见性和不可预测性，收集能体现异常情况的数据颇具挑战性。因此，大多数研究聚焦于无监督方法，即仅使用无缺陷样本（也称为标称样本）训练的算法。当前，多数现有的异常检测方法侧重于分析RGB图像。然而，在许多工业场景中，仅基于彩色图像难以有效识别异常，例如，光照条件变化易导致误检，表面偏差可能不会以明显异常的颜色呈现。结合彩色图像和三维传感器获取的表面信息可解决上述问题，并显著提升异常检测效果。
CVPR 2024 | 基于跨模态特征映射的多模态工业异常检测
m0_59235699的博客
06-17 1758
最近的研究表明，利用点云和图像来定位异常具有很大的潜力。然而，这些方法在工业制造中的应用往往受到显著的限制，例如使用内存库会导致内存占用和推理时间的大幅增加。作者提出了一种新颖的轻量级快速框架，该框架通过学习在正常样本上将一种模态的特征映射到另一种模态，并通过观察到的特征与映射特征之间的不一致性来检测异常。大量实验表明，该方法在MVTec 3D-AD数据集上实现了最先进的检测和分割性能，并且在标准和小样本设置下均表现出色，同时推理速度更快，内存占用更少。此外，作者还提出了一种层剪枝技术，以在性能损失较小的情
《Multimodal Industrial Anomaly Detection via Hybrid Fusion》论文学习及代码复现
m0_58520624的博客
11-03 3322
基于2D的工业异常检测已经得到了广泛的讨论研究，而基于3D点云和RGB图像的多模态工业异常检测仍然有许多未涉足的领域，现有的多模态工业缺陷检测方法直接将多模态特征连接起来，导致不同特征之间存在干扰，影响检测性能。3、DLP（决策层融合）：将多模态信息和多个记忆库相结合，并使用2个可学习模块Da和Ds进行检测与分割，其中Mrgb、Mfs、Mpt是记忆库， Φ和ψ是单个记忆库检测和分割的得分函数、P是记忆库构建算法。2、基于预训练特征提取器的方法：将提取到的特征映射到正态分布，然后将分布外的特征作为异常。
CVPR2023检测相关Detection论文速览上
weixin_44287798的博客
06-22 1960
CVPR2023论文速览Detection上
每日学术速递3.3
与君共勉，一起学习
03-03 660
标题：使用 3D 感知合成数据对人脸识别进行姿势影响估计并将其应用于质量评估作者：Marcel Grimmer, Christian Rathgeb, Christoph Busch文章链接：https://arxiv.org/abs/2303.00491v1项目代码：https://github.com/datasciencegrimmer/syn-yawpitch摘要： 评估面部图像的质量对于以足够的准确性操作面部识别系统至关重要。人脸质量标准化的最新进展 (ISO/IEC WD 2979
Multimodal Industrial Anomaly Detection by Crossmodal Reverse Distillation
ZHW-鲜橙大AI课题组的博客
01-06 1800
笔记作者：乔幸荣，女，西安工程大学电子信息学院，2024级研究生，张宏伟人工智能课题组研究方向：模式识别与智能系统电子邮件：2029518801@qq.com这篇文章针对现有的基于知识蒸馏的多模态异常检测方法可能无法有效捕获某一模态中的异常，从而导致检测失败以及不能充分利用丰富的内部和跨模式信息，提出了基于多分支蒸馏的跨模态反蒸馏方法。
windows下面的SpinNet实现出现的一丢丢问题
weixin_48818065的博客
05-05 2165
windows跑SpinNet出现的问题记录
PointNet++ pytorch版本中pointnet2_ops_lib的配置
热门推荐
qq_40321547的博客
04-23 1万+
PointNet++ pytorch版本中pointnet2_ops_lib的配置
把 cc-switch 搬上鸿蒙 PC 的那些天：一个 Tauri 应用的移植笔记
猫哥 的沉淀、积累、总结。天天学习，好好向上...c/c++,嵌入式 linux,Android,HarmonyOS，AIOT)
08-18 361
本文记录了将Tauri应用移植到鸿蒙PC的真实过程。作者选取GitHub热门项目cc-switch（12万+星标）作为移植对象，该工具采用Tauri v2架构（React+Rust）。移植面临三大挑战：1)依赖解析冲突需通过patch机制解决；2)交叉编译问题包括openssl、rquickjs等库的适配；3)构建系统细节如Windows环境下的manifest误判。最终方案基于社区OpenHarmony版Tauri fork，通过NAPI桥接+ArkWeb渲染实现。文章特别指出鸿蒙开发需要NDK与完整SD
AI Agent学习笔记（20260817）
weixin_43341767的博客
08-18 240
肖斌。
【WMS学习笔记系列】02-系统架构设计
IT古董
08-18 406
本文介绍了WMS（仓库管理系统）的系统架构设计，重点推荐采用模块化单体架构（Modular Monolith）作为中小型WMS的初始架构方案。该架构具有开发效率高、部署运维简单等优势，适合团队规模小、业务初期的场景。文章详细说明了技术栈选型（MySQL、Redis、RabbitMQ等）、模块划分方案（核心模块、入库/出库等业务模块），以及包结构规范。同时提出了库存扣减、库位推荐、波次策略等核心业务设计，并阐述了分布式锁的应用场景和安全架构设计（RBAC权限模型）。该架构支持未来平滑过渡到微服务架构，为系统演
大模型并行推理架构笔记
谢白羽
08-19 306
摘要： 大模型推理面临显存和计算瓶颈，需通过并行策略优化。核心方法分为两类：模型并行（TP、PP、EP）和请求并行（批处理优化）。TP按层内切分降低单请求延迟，PP跨层切分减少通信量，EP专用于MoE模型。连续批处理和PagedAttention提升吞吐效率。进阶策略包括PD分离（Prefill与Decode阶段解耦）、序列并行（长上下文优化）和投机解码（小模型加速生成）。近两年进展聚焦分离式架构（如AF分离）和硬件超节点（如GB200 NVL72），以更高粒度切分计算与带宽负载。选型需权衡规模、延迟需求与
PostgreSQL笔记9: psql 客户端——从入门到精通
Wang的专栏
08-17 612
PostgreSQL命令行工具psql是一个强大的交互式终端，集SQL执行、元数据查看和系统命令桥接于一体。文章系统介绍了psql的三种使用模式（交互式、脚本执行、命令行参数）、丰富的元命令体系（如对象查看命令\d家族、格式化控制\x、执行计时\timing等）、实用命令行选项（-c执行SQL、-f执行脚本）以及个性化配置方法（psqlrc启动文件、提示符定制）。此外，还分享了高级技巧如动态SQL执行\gexec、外部编辑器集成\e等，帮助用户提升数据库管理和开发效率。
C语言指针与内存核心笔记：从底层原理到避坑指南
最新发布
carlos20031004的博客
08-20 511
本笔记系统梳理了C语言指针与内存管理的核心知识。内容涵盖指针的本质、解引用及内存地址概念，深度剖析值传递与指针传递在变量交换中的作用域差异。重点解析了sizeof运算陷阱、数组与指针的等价关系及二维数组的内存布局。同时，详细讲解了malloc/free动态内存分配机制、二级指针模型及函数传参规范，并总结了野指针、内存泄漏等常见易错点，旨在帮助开发者夯实底层基础，编写安全高效的代码。
linux笔记归纳15：自定义协议与序列化
yonakaame的博客
08-18 273
目录自定义协议与序列化一、应用层1.1.应用层协议1.2.序列化和反序列化1.3.重新理解read和write二、Jsoncpp2.1.Jsoncpp概念2.2.Jsoncpp安装2.3.序列化2.4.反序列化2.5.手写序列化和反序列化三、进程间关系与守护进程3.1.进程组3.2.组长进程3.3.会话3.4.任务3.5.控制终端3.6.作业控制3.7.守护进程3.8.setsid函数3.9.null文件3.10.daemon函数四、网络版计算器2.1.socket封装2.2.自定义协议2.3.设计服务器。
评论 5
添加红包

请填写红包祝福语或标题

个

红包个数最小为10个

元

红包金额最低5元

当前余额3.43元 前往充值 >
需支付：10.00元
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付元
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明：

1.余额是钱包充值的虚拟货币，按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载，可以购买VIP、付费专栏及课程。

余额充值