Neural Network: Sample Code for caffe feature Visualization

Sample Code for caffe feature Visualization

Many readers read the previous two articles

Summarize the research process of using caffe to run image data.

Summary of deep learning practical experience 2-accuracy improved again, reaching 0.8.

Then, I want to know how to implement feature visualization.

To put it simply, it is to let the neural network spread forward once, then extract the feature values of a layer, and convert them to images for storage.

Below is a demo that you can modify as needed.

Let's take a look at the usage of my demo.

Visualize_features.bin net_proto pretrained_net_proto iterations [CPU/GPU] img_list_file dstdir laydepth

Visualize_features.bin is an executable file compiled by CPP.

Let's take a look at the meaning of each parameter:

1 net_proto: A file format defined by caffe that defines the network structure. The suffix is ". prototxt ". This file defines the network input, related parameters, and the overall network structure.

2 pretrained_net_proto: This is a trained model.

3 iterations: number of iterations

4 [CPU/GPU]: CPU or GPU Mode

5 img_list_file: List of file names to be tested. Here I need this mainly to get the Class Name of the image.

6 dstdir: the output folder of the image.

7 laydepth: the layer of features to be output

The following is an example:

. /Visualize_features.bin/home/linger/Caffe-action/Caffe-master/examples/cifar10/examples/home/linger/Caffe-action/Caffe-master/examples/cifar10 /cifar10_full_iter_60000 20 GPU/home/linger/testfile/skirt_test_attachment/image_filename/home/linger/testfile/innerproduct/7

The source code is as follows:

// Copyright 2013 yangqing Jia // This is a simple script that allows one to quickly test a network whose // structure is specified by text format protocol buffers, and whose parameter // are loaded from a pre-trained network. // usage: // test_net net_proto pretrained_net_proto iterations [CPU/GPU] # include <cuda_runtime.h> # include <fstream> # include <iostream> # include <cstring> # include <cstdlib> # inclu De <algorithm> # include <vector> # include <utility> # include "caffe/Caffe. HPP "# include <opencv2/highgui. HPP> # include <opencv2/highgui/highgui_c.h> # include <opencv2/imgproc. HPP> Using STD: make_pair; Using STD: pair; using namespace caffe; // nolint (build/namespaces) using namespace STD; vector <string> filenames; char * filelist; /** the content format of the file to be read is similar to this: global ID category _ category id.jpg 0 One-Step skirt _0.jpg 1 one-step skirt _1.jpg 2 One-step podium _10.jpg */void readfile () {If (filenames. empty () {ifstream read (filelist ); // "/home/linger/testfile/test_attachment/image_filename"/"/home/linger/imdata/test_files_collar.txt" // "/home/linger/testfile/ testfilename "If (read. is_open () {While (! Read. EOF () {string name; int ID; read >>> ID >>> name; filenames. push_back (name) ;}}}/ ** obtain the class name */string getclassnamebyid (int id) {readfile (); int Index = filenames [ID] based on the image ID. find_last_of ('_'); Return filenames [ID]. substr (0, index);} void writebatch (const float * data, int num, int channels, int width, int height, int startid, const char * DIR) {for (INT id = 0; id <num; Id ++) {for (INT channel = 0; Channel <channels; Channel ++) {CV :: mat MAT (height, width, cv_8uc1); // high width vector <float> VEC; Vec. resize (height); float max =-1; float min = 999999; For (int row = 0; row