Summary of CNN Visualization Technology (I)- Visualization of feature map
Summary of CNN visualization Technology (II) — Visualization of convolution kernel
Summary of CNN Visualization Technology (III) — Class visualization
Welcome to the public account CV technical guide, focusing on the technical summary of computer vision, the latest technology tracking, classical paper interpretation.
Introduction:
Three visualization methods, feature map visualization, convolution kernel visualization and class visualization, have been introduced previously. These three methods are very common in many papers proposing new models or new methods. Their main function is to improve the credibility of the model or new method, or to increase the workload, or to make up the number of words. Others help to understand how the model learns for a particular task, what information is learned, and what areas are affected by recognition.
This article will introduce some visual projects, including CNN interpreter, feature map, convolution kernel, some code and projects for class visualization, structure visualization tools, and network structure manual drawing tools.
1. CNN-Explainer
This is an online interactive visualization tool called CNN Interpreter published by a Chinese doctor. Are mainly for the beginners of deep learning small white people understand how to work on the neural network is helpful, such as convolution process, ReLU process, the average pooling process, each layer between the appearance of the characteristics of the figure, you can see, equivalent to a microscope, you can casually to any one, any changes before and after an operation, observe clearly.
Show the changes of the feature map before and after the convolution process, and the operation in the middle.
How does CNN output predictions
You can also upload pictures online to see the changes of an image after each layer of convolution, pooling and activation, and finally output the prediction results.
Project link:
Github.com/poloclub/cn…
2. Some visual feature maps, convolution kernel, heat map code.
Visual feature map
Github.com/waallf/Vius…
Visual convolution kernel
Keras. IO/examples/vi…
Blog. Keras. IO/how – convolu…
Grad-CAM
Github.com/ramprs/grad…
Heat map
Github.com/heuritech/c…
The following project is a link that includes both feature map visualization, convolution kernel visualization and heat map:
Github.com/raghakot/ke…
3. Structure visualization tools
Netscope
An online tool for visualizing model structures, only supports visualizing Caffe’s Prototxt file. You need to write your own prototxt file.
This picture is from the network, invaded and deleted
Project Address:
Github.com/ethereon/ne…
ConvNetDraw
The tool is illustrated directly with two diagrams, the first for input and the second for output
These two images are from the network
Project Address:
Github.com/cbovar/Conv…
PlotNeuralNet
This is a little bit more troublesome, and the renderings are as follows:
Project Address:
Github.com/HarisIqbal8…
4. Manual network structure drawing tool
One of the questions many novices ask is how the network structures in the paper were drawn.
Here is the answer. What I know is mainly using POWERPOINT, VISIO. You can also use the ones above.
Add an online tool, NN-SVG
Alexlenail. me/ nn-svg /
conclusion
These four articles have basically introduced some visualization methods of CNN at present, namely feature map visualization, convolution kernel visualization and class visualization, and summarized some visualization tools and projects. Of course, there are some missing ones. In the future, if there are some visualization tools with major breakthroughs, we will continue to supplement them. It will be placed in the technical summary section of the public account (CV technical Guide).
As for visualization, it also includes visualization of training process, such as real-time update of Loss value and precision, which is relatively simple and will not be explained in this summary series.
This article comes from the technical summary series of the public CV technical guide.
Welcome to the public account CV technical guide, focusing on the technical summary of computer vision, the latest technology tracking, classical paper interpretation.
To get a PDF summary of the following articles, reply to the official account with the keyword “Technical summary”.
Other articles
Shi Baixin, Peking University: From the reviewer’s perspective, how to write a CVPR paper
Siamese network summary
Summary of computer vision terms (1) to build a knowledge system of computer vision
Summary of underfitting and overfitting techniques
Summary of normalization methods
Summary of common ideas of paper innovation
Summary of methods of reading English literature efficiently in CV direction
A review of small sample learning in computer vision
A brief overview of knowledge distillation
Optimize OpenCV video read speed
NMS summary
Technical summary of loss function
Technical summary of attention mechanisms
Summary of feature pyramid technology
Summary of pooling technology
Summary of data enhancement methods
Summary of CNN structure Evolution (I) Classic model
Summary of CNN structure evolution (II) Lightweight model
Summary of CNN structure evolution (III) Design principles
How to view the future of computer vision
Summary of CNN Visualization Technology (I) – Visualization of feature map
CNN Visualization Technology Summary (II) – convolution kernel visualization
Summary of CNN Visualization Technology (III) – Class visualization
CNN Visualization Technology Summary (IV) – Visualization tools and projects