NJUyued/NoW

PC2-NoiseofWeb

This repo releases data introduced in our paper accepted:

PC2: Pseudo-Classification Based Pseudo-Captioning for Noisy Correspondence Learning in Cross-Modal Retrieval
Authors: Yue Duan, Zhangxuan Gu, Zhenzhe Ying, Lei Qi, Changhua Meng and Yinghuan Shi

• Quick links: [Code | [PDF/Abs-arXiv | PDF/Abs-Published (coming soon) | Slides/Video (coming soon) | 文章解读-知乎(Zhihu) | 视频解读-bilibili | Dataset download]

• 📰 Latest news:

  • We provide a video presentation (in chinese) of this work on bilibili.
  • We write a detailed explanation (in chinese) of this work on Zhihu.
  • Our paper is accepted by ACM International Conference on Multimedia (ACM MM) 2024 🎉🎉. Thanks to users.

Data Collection

We develop a new dataset named Noise of Web (NoW) for NCL. It contains 100K image-text pairs consisting of website images and multilingual website meta-descriptions (98,000 pairs for training, 1,000 for validation, and 1,000 for testing). NoW has two main characteristics: without human annotations and the noisy pairs are naturally captured. The source image data of NoW is obtained by taking screenshots when accessing web pages on mobile user interface (MUI) with 720 X 1280 resolution, and we parse the meta-description field in the HTML source code as the captions. In NCR (predecessor of NCL), each image in all datasets were preprocessed using Faster-RCNN detector provided by Bottom-up Attention Model to generate 36 region proposals, and each proposal was encoded as a 2048-dimensional feature. Thus, following NCR, we release our the features instead of raw images for fair comparison. However, we can not just use detection methods like Faster-RCNN to extract image features since it is trained on real-world animals and objects on MS-COCO. To tackle this, we adapt APT as the detection model since it is trained on MUI data. Then, we capture the 768-dimensional features of top 36 objects for one image. Due to the automated and non-human curated data collection process, the noise in NoW is highly authentic and intrinsic. The estimated noise ratio of this dataset is nearly 70%.

Data Structure

|-- h5100k_precomp
|   |-- dev_caps_bpe.txt
|   |-- dev_caps_bert.txt
|   |-- dev_caps_jieba.txt
|   |-- dev_ids.txt
|   |-- dev_ims.npy
|   |-- test_caps_bpe.txt
|   |-- test_caps_bert.txt
|   |-- test_caps_jieba.txt
|   |-- test_ids.txt
|   |-- test_ims.npy
|   |-- train_caps_bpe.txt
|   |-- train_caps_bert.txt
|   |-- train_caps_jieba.txt
|   |-- train_ids.txt
|   |-- train_ims.npy
|-- vocab
|   |-- now100k_precomp_vocab_bert.json
|   |-- now100k_precomp_vocab_bpe.json
|   |-- now100k_precomp_vocab_jieba.json

Please note that since our raw data contains some sensitive business data, we only provide the encoded image features (*_ims.npy) and the token ids of the text tokenized. For tokenizer, we provide Tokenizers with BPE to produce *_caps_bpe.txt, BertTokenizer with bert-base-multilingual-cased pre-trained model to produce *_caps_bert.txt, and Jieba to produce *_caps_jieba.txt. Our vocabulary size of BPETokenizer is 10,000, while BertTokenizer and JiebaTokenizer have a vocabulary size of 32,702 and 56,271 respectively. (recorded in now100k_precomp_vocab_*.txt). *_ids.txt records the data indexs in the original 500k dataset. In the future, we may process and make the original dataset public.

Usage

# data_path: your dataset name and path
# data_split: {train,dev,test}
# tokenizer: {bpe,bert,jieba}
# vocabulary size of {bpe,bert,jieba} is {10000,32702,56271} 

# captions
with open(os.path.join(data_path, "{}_caps_{}.txt".format(data_split, tokenizer))) as f:
    for line in f:
        captions.append(line.strip())
captions_token = []
for index in range(len(captions)):
  caption = captions[index]
  tokens = caption.split(',')
  caption = []
  caption.append(vocab("<start>"))
  caption.extend([int(token) for token in tokens if token])
  caption.append(vocab("<end>"))
  captions_token.append(caption)

# images
images = np.load(os.path.join(data_path, "%s_ims.npy" % data_split))

return captions_token, images

Additionally, you can search for code snippets containing the string now100k_precomp in co_train.py, data.py, evaluation.py, and run.py in PC2's repo and refer to them to process the NoW dataset for use in your own code.