yaoyao-liu
commented
5 years ago Hi Li,
Thanks for your attention on this project.
DATA_LIST1 and DATA_LIST2 are not the real filenames. If you want to run the code, you should replace them with the directories of your downloaded data list files. For example, If you download the splits in your current folder, the command must be changed to:
python main.py --train=True --net_arch=vgg19 --double_stream_mode=True --epoch_num=10 --batch_size=10 --cls_num=16 --learning_rate=1e-4 --shuffle_dataset=True --img_list1=./train_test_eval_splits/relation_consistency3/single_body1_train_16.txt --img_list2=./train_test_eval_splits/relation_consistency3/single_body1_train_16.txt
As for how to choose the datalist, you may read the paper provided on the GitHub and this project page: https://www.mpi-inf.mpg.de/departments/computer-vision-and-multimodal-computing/research/human-activity-recognition/social-relation-recognition/
If you have any further questions, feel free to contact me. You may directly send me email: yaoyaoliu@outlook.com
noticeable
hello thx for you share of your code.I'm going to cite your parper for a further reacher.however, when I use the comman: python main.py --train = True --net_arch = vgg19 --double_stream_mode = True --epoch_num = 10 --batch_size = 10 --cls_num = 16 --learning_rate = 1e-4 --shuffle_dataset = True --img_list1 = $ {DATA_LIST1} --img_list2 = $ {DATA_LIST2} to train the net. there is show a error about DATA_LIST1 and DATA_LIST2.I have download all resource you said in Readme.and I have tried to use like : python main.py --train=True --net_arch=vgg19 --double_stream_mode=True --epoch_num=10 --batch_size=10 --cls_num=16 --learning_rate=1e-4 --shuffle_dataset=True --img_list1=${./all_single_face} --img_list2=${./all_single_body}
but it show errors too.could you please tell what's mean is it?
Looking forward to your reply.