Weize Li

作者您好， 首先感谢您的代码贡献，非常简洁，关键注释非常清晰！已按照readme已经成功跑通示例~目前希望依托您的代码框架，进一步想试一试引入自己的预训练网络，生成gradcam，进行图像异常检测。现有一个基于Res18预训练模型，前面添加了head，后面添加了几层额外的卷积层和fc层（最终输出分别是0：正常、1：异常），对自己的正常数据集进行无监督学习训练得到权重，用于对异常图片进行异常检测。然后利用gradcam在异常图像上标注出异常的位置。 现有问题是如何将前面提到的自己的模型权重引入框架？自己试了试之后会有如下报错，请问可能是什么问题？ **(gradcam)_____@server3090-X570-AORUS-PRO-WIFI:~/Grad-CAM.pytorch-master$ python main.py** feature shape:torch.Size([1, 512, 7, 7]) /home/____/.conda/envs/gradcam/lib/python3.8/site-packages/torch/nn/modules/module.py:1033: UserWarning: Using a non-full backward hook when the forward contains multiple autograd Nodes is deprecated and will be...

Editing shapes in implicit space is really cool, and I really want to try the demo! Unfortunately, I got everything fixed, but I encountered this problem:  (In Linux env)...

Great work and thanks for open source! There's one thing about the create env stage, could you please check your requirements.txt content? I found that txt. content does not support...