ind_knn_ad环境搭建和运行【用自己的数据集】

源码

https://github.com/rvorias/ind_knn_ad

conda create -n padim python=3.7
unzip ind_knn_ad-master.zip
cd autodl-tmp/ind_knn_adconda activate padimpip install -r requirements.txtpip install click
pip install numpy

修改数据集根路径

# data.py
DATASETS_PATH = Path(r"/root/autodl-tmp/SimpleNet-main/data4")

加载测试集需要考虑后缀名兼容性

    def __getitem__(self, index):path, _ = self.samples[index]sample = self.loader(path)if "good" in path:target = Image.new("L", (self.size, self.size))sample_class = 0else:target_path = path.replace("test", "ground_truth")
#             target_path = target_path.replace(".png", "_mask.png")# 提取文件名和扩展名filename, file_extension = os.path.splitext(os.path.basename(target_path))# 生成掩码文件名mask_filename = f"{filename}_mask.png"# 生成掩码文件路径target_path = os.path.join(os.path.dirname(target_path), mask_filename)

删除或跳过测试点文件

因为他是调用的父类的构造函数，所以直接删掉了测试点文件夹

问题huggingface_hub.errors.LocalEntryNotFoundError:

https://blog.csdn.net/qq_43920838/article/details/144508431
需要先配置镜像：
export HF_ENDPOINT=“https://hf-mirror.com”

python run.py spade --dataset bottle