search

tswang0116 / TA-DCH

Source code for paper "Targeted Adversarial Attack for Deep Cross-modal Hashing Retrieval".
13 stars 4 forks source link

您好，请问您这篇文章在哪里可以搜到呢 #2

Closed zzidlezz closed 1 year ago

tswang0116 commented 2 years ago

目前还在投稿，录用后将会公开

zzidlezz commented 2 years ago

请问您的IAPR-TC12这个数据集怎么打开呢，我尝试了matlab不能打开，python的f5py的包按照您load_data里的 images = Data['IAll'][:] labels = Data['LAll'][:] tags = Data['TAll'][:]这样也打不开

tswang0116 commented 2 years ago 
Data = h5py.File(’IAPR-TC12数据集的存放路径‘)
images = Data['IAll'][:]
labels = Data['LAll'][:]
tags = Data['TAll'][:]

如果这样打不开，麻烦把您的错误写一下

zzidlezz commented 2 years ago

可以读取了，我自己的问题，您的这些图片224x224是已经预处理过了吗

tswang0116 commented 2 years ago

没有经过其他处理，只是把图像放缩到224*224了

zzidlezz commented 1 year ago

您好，请问您提供的例子里那个DGCPN.pth是用DGCPN模型在本文提供的数据集上训练出来的模型，然后加载到这个代码中的吗，其他攻击的模型也是同理吧

tswang0116 commented 1 year ago

六种被攻击的跨模态哈希检索方法均在我们所提供的数据集上训练得到，然后按照说明放到相应文件夹中即可

zzidlezz commented 1 year ago

您好这篇文章最后可视化的结果，就是给定查询的文本返回前几张图片的代码可以参考下吗

tswang0116 commented 1 year ago

def calc_hamming_dist(B1, B2): q = B2.shape[1] if len(B1.shape) < 2: B1 = B1.unsqueeze(0) distH = 0.5 * (q - B1.mm(B2.transpose(0, 1))) return distH

def Result(qB, rB, query_label, retrieval_label, k): num_query = query_label.shape[0] index = torch.zeros(num_query, k) for i in range(num_query): hamm = calc_hammingdist(qB[i, :], rB) , ind = torch.sort(hamm) ind.squeeze_() index[i] = ind[:k] return index