rbgirshick / py-faster-rcnn

Faster R-CNN (Python implementation) -- see https://github.com/ShaoqingRen/faster_rcnn for the official MATLAB version
Other
8.07k stars 4.11k forks source link

Issue with coco training #132

Open kaishijeng opened 8 years ago

kaishijeng commented 8 years ago

I follow the instruction in faster_rcnn_end2end.sh and use the following command for a coco traning.

./experiments/scripts/faster_rcnn_end2end.sh 0 VGG_CNN_M_1024 pascal_voc \ --set EXP_DIR foobar RNG_SEED 42 TRAIN.SCALES "[400, 500, 600, 700]"

However there is an assert len(cfg_list) % 2 == 0, from the following function: If I remove TRAIN.SCALES "[400, 500, 600, 700]", then the assert is gone. Do you know what is thr purpose TRAIN.SCALES "[400, 500, 600, 700]" and why there is an assert?

def cfg_from_list(cfg_list): """Set config keys via list (e.g., from command line).""" from ast import literal_eval assert len(cfg_list) % 2 == 0

hyichao commented 8 years ago

I met the same problem trying to train pascal_voc. I try google the error but nothing found. If somebody know how to use the script with correct parameters and why, please tell

faschinj commented 8 years ago

Make sure there are no spaces after the commas otherwise they are parsed as separate arguments like so TRAIN.SCALES "[400,500,600,700]"

sunformoon commented 7 years ago

@faschinj Thank you, solved this assertion.

zhangxujinsh commented 7 years ago

@faschinj Thank you, solved this assertion.