caffe绘制loss和accuracy曲线

自带工具

1、训练时后面加个参数

$TOOLS/caffe train --solver=$SOLVERFILE 2>&1 |tee out.log

2、在tools/extra目录下有个parse_log.py，用它来解析日志文件

python parse_log.py out.log ./ #两个参数，一个是日志文件，另一个是保存的路径

运行后会生成两个文件out.log.train和out.log.test

3、生成图片

./plot_training_log.py.example 0  save.png out.log //注意这个是out.log

Notes:  
    1. Supporting multiple logs.  
    2. Log file name must end with the lower-cased ".log".  
Supported chart types:  //可以修改数字来绘制不同曲线
    0: Test accuracy  vs. Iters  
    1: Test accuracy  vs. Seconds  
    2: Test loss  vs. Iters  
    3: Test loss  vs. Seconds  
    4: Train learning rate  vs. Iters  
    5: Train learning rate  vs. Seconds  
    6: Train loss  vs. Iters  
    7: Train loss  vs. Seconds