jonneslin/evison

We provide an easy way for visualizing

easy_visualization

We provide an easy way for visualizing

You will get a heatmap image when you send it the layer name, the model and the image.

How

the example code is located in the root, its name is 'Evison_Tutorial.ipynb'

Online

We provide an example on goolge colab: https://colab.research.google.com/drive/1nNxHOur741WMjBIp1BHrmGi9xHb_CfDB?usp=sharing

Multi-output

You may specify which output you want to use for classification bythe argument target_output.

display = Display(network, visualized_layer, target_output=2, img_size=(224, 224)) # You may use the output of index 3 for classification.

Reference

https://github.com/utkuozbulak/pytorch-cnn-visualizations

Contact

Jones(Jinhong) Lin: [email protected] / [email protected]

evison's Issues

能提高pytorch的版本吗?

该如何解决

IndexError: list index out of range

RuntimeError: The size of tensor a (4) must match the size of tensor b (3) at non-singleton dimension 0

如何解决？

AttributeError: 'tuple' object has no attribute 'cpu' 如何解决？

作者你好，在之前的issue里面作者说已解决该问题并已经修改上传至github上，然而并没有，烦请作者更新或者给出问题解决方案。出现的错误信息与前值一致，即：AttributeError: 'tuple' object has no attribute 'cpu' 。

您好，该方法可以用在分割网络的特征图上么

您好，该方法可以用在分割网络的特征图上么

如何直接在运行栏中呈现效果图？

请问如何直接在jupyter notebook中运行直接呈现效果图呢，而不是保存在目录下的文件夹中，这样观察起来不太方便...求解答！

hidden feature

hello,i input an image with shape(3,480,480), i want to vision the hidden feature (240,240).
there is an error about numpy's broadcast .

cam += w.cpu().data.numpy() * feature_map[i, :, :].cpu().data.numpy()
ValueError: operands could not be broadcast together with shapes (480,480) (240,240)

TypeError: forward() missing 1 required positional argument: 'img'

请问这个错误是什么原因？看源代码中没有img这个变量啊

AttributeError: 'tuple' object has no attribute 'cpu' 如何解决？

if target_class is None:
target_class = np.argmax(output.cpu().data.numpy())
好像是Evision里生成cam时候报的错， feature_map, output = self.forward(image),是不是output没有cpu属性？

moco 该如何修改进行特征可视化

你好，请问 moco 系列的模型该如何修改进行特征可视化？谢谢啦