CAFFE-SSD中loss、accuracy曲线可视化（双Y轴）

阿呀么耀博主的博客可以完成loss、acc画在一张表里，我参考此博客并改进了一下，可以画出双Y轴图表。

1.利用输出的.log文件（应该在caffe/jobs/VGGNet/VOC0712/SSD_300x300目录下）生成.test .train文件。

cd到caffe/tools/extra下

./parse_log.sh ../../jobs/VGGNet/VOC0712/SSD_300x300/VGG_VOC0712_SSD_300x300.log

生成以下四个文件

打开一个看一下

2.下面就是重点，画双Y轴loss、acc曲线。

不多哔哔，代码如下

# -*- coding: utf-8 -*-
#!/usr/bin/env python
import numpy as np
import matplotlib.pyplot as plt
l_n = 0
a_n = 0
loss_x = []
loss_y = []
acc_x = []
acc_y = []
with open("/home/shadiao/caffe/python/可视化/VGG_VOC0712_SSD_300x300.log.train") as loss:   #根据自己的目录做对应的修改
    for line in loss:
        #line = line.strip()
        #print(line.split()[0])
        if float(line.split()[0]) == l_n:
            loss_y.append(float(line.split()[2]))  
            loss_x.append(l_n)
            l_n += 1000   #根据display调整，如果太密可以设置为display的倍数
with open("/home/shadiao/caffe/python/可视化/VGG_VOC0712_SSD_300x300.log.test") as acc:   #根据自己的目录做对应的修改
    for line in acc:
        #line = line.strip()
        if float(line.split()[0]) == a_n:
            acc_y.append(float(line.split()[2]))
            acc_x.append(a_n)
            a_n += 10000   #根据test_interval调整
fig = plt.figure(figsize=(8,6))
ax = fig.add_subplot(111)
ax.plot(loss_x,loss_y,'',label="loss")
ax.set_title("loss and acc vs iter")

ax2 = ax.twinx()
ax2.plot(acc_x,acc_y,'r',label="acc")
fig.legend(bbox_to_anchor=(1,0.9), bbox_transform=ax.transAxes)
ax.set_xlabel('iter')
ax.set_ylabel(r"loss")
ax2.set_ylabel(r"acc")

plt.savefig('train_loss.png')

我画的结果图如下

reference

[1]https://blog.csdn.net/CAU_Ayao/article/details/84502826
[2]https://blog.csdn.net/qq_36219202/article/details/78338920