Plot 4 loss curves using 'parse_log.py' tool

[flipflop98]

Hello every py-faster-rcnn users,

Here I write one error which occurs while running 'parse_log.py' program. Any idea?

Thank you for any comments in advance.

ch723.kim@ubuntu:/samba/py-faster-rcnn-master$ python ./caffe-fast-rcnn/tools/extra/parse_log.py ./experiments/logs/kitti_faster_rcnn_ZF.txt.2016-01-19_21-32-39 .
Traceback (most recent call last):
File "./caffe-fast-rcnn/tools/extra/parse_log.py", line 209, in
main()
File "./caffe-fast-rcnn/tools/extra/parse_log.py", line 202, in main
train_dict_list, test_dict_list = parse_log(args.logfile_path)
File "./caffe-fast-rcnn/tools/extra/parse_log.py", line 50, in parse_log
start_time = extract_seconds.get_start_time(f, logfile_year)
File "/home/ch723.kim/samba/py-faster-rcnn-master/caffe-fast-rcnn/tools/extra/extract_seconds.py", line 47, in get_start_time
start_datetime = extract_datetime_from_line(line, year)
File "/home/ch723.kim/samba/py-faster-rcnn-master/caffe-fast-rcnn/tools/extra/extract_seconds.py", line 10, in extract_datetime_from_line
month = int(line[0][1:3])
ValueError: invalid literal for int() with base 10: 'ol'
ch723.kim@ubuntu:/samba/py-faster-rcnn-master$

[Sancerio]

Hi @flipflop98, the parse_log.py contains code that looks for the string "Iteration" for every line and assume that the month (Expected format: I0210 13:39:22.381027 25210 solver.cpp:204] Iteration 100, lr = 0.00992565) is always embedded in that line. You can verify that this is not the case in the faster-rcnn log file by searching for "Iteration" in it. You can modify the code to only extract_seconds if the line is started with I. Hope this helps!

[leon-woo]

I've modified the parse_log.py for faster_rcnn, but only for this project. Hers are my files, https://pan.baidu.com/s/1slaVUzV, just copy these files to caffe-fast-rcnn/tools/extra/, and cover the original corresponding files.

[peterWon]

@leon-woo Hi ,could you please tell me how to generate the log file? I can't find it in my 'tmp' folder. Thanks a lot!

[leon-woo]

@peterWon you can use it just as the original version in caffe, but only for 4 step training log:)

[lolongcovas]

thanks @leon-woo for sharing new plot. however, the link it seems broken, can u paste the code here? best