使用Detectron2预训练模型时触发NotImplementedError问题求助

执行以下代码初始化Detectron2的DefaultPredictor时出现NotImplementedError：

安装与导入代码 #

import torch
torch.__version__
import torchvision
#torchvision.__version__
!pip install detectron2 -f https://dl.fbaipublicfiles.com/detectron2/wheels/cu102/torch1.7/index.html

import detectron2
from detectron2.utils.logger import setup_logger
setup_logger()

import numpy as np
import os, json, cv2, random
import matplotlib.pyplot as plt
%matplotlib inline
from detectron2 import model_zoo
from detectron2.engine import DefaultPredictor
from detectron2.config import get_cfg
from detectron2.utils.visualizer import Visualizer
from detectron2.data import MetadataCatalog, DatasetCatalog
from detectron2.structures import BoxMode

配置与初始化代码 #

cfg = get_cfg()
cfg.merge_from_file(model_zoo.get_config_file("COCOInstanceSegmentation/mask_rcnn_R_50_FPN_3x.yaml"))
cfg.MODEL.ROI_HEADS.SCORE_THRESH_TEST = 0.5 
cfg.MODEL.WEIGHTS = model_zoo.get_checkpoint_url("COCOInstanceSegmentation/mask_rcnn_R_50_FPN_3x.yaml")

predictor = DefaultPredictor(cfg)

错误信息 #

---------------------------------------------------------------------------
NotImplementedError                       Traceback (most recent call last)
<ipython-input-27-699e754fc9df> in <module>
----> 1 predictor = DefaultPredictor(cfg)

4 frames
/usr/local/lib/python3.8/dist-packages/iopath/common/file_io.py in _isfile(self, path, **kwargs)
    438             bool: true if the path is a file
    439         """
--> 440         raise NotImplementedError()
    441 
    442     def _isdir(self, path: str, **kwargs: Any) -> bool:

NotImplementedError: 

解决方法 #

• 升级iopath库：该错误通常由iopath版本不兼容导致，执行以下命令升级：

  pip install --upgrade iopath
  

• 确保版本匹配：检查PyTorch、CUDA与Detectron2版本是否对应。当前使用的是CUDA10.2+PyTorch1.7，需确保安装的Detectron2轮子适配该环境，版本不匹配会引发依赖问题。

• 手动加载本地权重：若自动下载权重出现问题，可手动下载后本地加载：

  1. 下载权重：

     wget https://dl.fbaipublicfiles.com/detectron2/COCOInstanceSegmentation/mask_rcnn_R_50_FPN_3x/137849600/model_final_f10217.pkl
     

  2. 修改配置中的权重路径：

     cfg.MODEL.WEIGHTS = "./model_final_f10217.pkl"
     

• 重新安装Detectron2：若上述方法无效，彻底卸载后重装对应版本：

  pip uninstall detectron2 -y
  pip install detectron2 -f https://dl.fbaipublicfiles.com/detectron2/wheels/cu102/torch1.7/index.html --force-reinstall
  

内容的提问来源于stack exchange，提问作者Keshav Khanal