treezzz/dhn_pytorch

Source code for paper "Deep Hashing Network for Efficient Similarity Retrieval" on AAAI-2016

dhn_pytorch's Issues

No results obtained using default parameters

Hello, I want to use DHN to verify palmprint recognition code. For this I use default parameters to train the DHN network, my settings are as follows:
parser = argparse.ArgumentParser(description='DHN_PyTorch')
parser.add_argument('--dataset', default='cifar-10',
help='Dataset name.')
parser.add_argument('--root', default='../cifar-10',
help='Path of dataset')
parser.add_argument('--code-length', type=int, default=128,
help='Binary hash code length.')
parser.add_argument('--arch', default='alexnet', type=str,
help='CNN model name.(default: alexnet)')
parser.add_argument('--batch-size', default=256, type=int,
help='Batch size.(default: 256)')
parser.add_argument('--lr', default=1e-2, type=float,
help='Learning rate.(default: 1e-5)')
parser.add_argument('--max-iter', default=500, type=int,
help='Number of iterations.(default: 500)')
parser.add_argument('--num-workers', default=6, type=int,
help='Number of loading data threads.(default: 6)')
parser.add_argument('--topk', default=-1, type=int,
help='Calculate map of top k.(default: all)')
parser.add_argument('--gpu', default=True, type=int,
help='Using gpu.(default: False)')
parser.add_argument('--lamda', default=1, type=float,
help='Hyper-parameter.(default: 1)')
parser.add_argument('--seed', default=3367, type=int,
help='Random seed.(default: 3367)')
parser.add_argument('--evaluate-interval', default=10, type=int,
help='Evaluation interval.(default: 10)')

But using this set of parameters, I can't get the result of MAP, and LOSS hasn't dropped. Is there any error in my verification method?

训练文件

您好！非常感谢您的开源，请问您可以分享一下nus_wide的train_img.txt和train_label_onehot_tc21.txt吗？谢谢！