Datasets:
Size:
100K<n<1M
ArXiv:
Tags:
image-text retrieval
noisy correspondence learning
NCL-specific benchmark
realistic
industry
mobile user interface
License:
Update README.md
Browse files
README.md
CHANGED
@@ -31,12 +31,12 @@ size_categories:
|
|
31 |
This repo releases data introduced in our paper
|
32 |
|
33 |
> ***PC2: Pseudo-Classification Based Pseudo-Captioning for Noisy Correspondence Learning in Cross-Modal Retrieval***
|
34 |
-
>
|
35 |
|
36 |
-
Quick links: [[Code](https://github.com/alipay/PC2-NoiseofWeb) | [
|
37 |
|
38 |
## Data Collection
|
39 |
-
We develop a new dataset named **Noise of Web (NoW)** for NCL. It contains **100K
|
40 |
|
41 |
<div align=center>
|
42 |
|
|
|
31 |
This repo releases data introduced in our paper
|
32 |
|
33 |
> ***PC2: Pseudo-Classification Based Pseudo-Captioning for Noisy Correspondence Learning in Cross-Modal Retrieval***
|
34 |
+
> **Authors**: ***[Yue Duan](https://njuyued.github.io/)**, Zhangxuan Gu, Zhenzhe Ying, Lei Qi, Changhua Meng and Yinghuan Shi*
|
35 |
|
36 |
+
Quick links: [[Code](https://github.com/alipay/PC2-NoiseofWeb) | [[PDF](https://arxiv.org/pdf/2408.01349)/[Abs](https://arxiv.org/abs/2408.01349)-arXiv | [Published paper (coming soon)]() | [Poster (coming soon)]() | [Zhihu](https://zhuanlan.zhihu.com/p/711149124) | [Dataset download](https://huggingface.co/datasets/NJUyued/NoW/resolve/main/NoW.zip?download=true)]
|
37 |
|
38 |
## Data Collection
|
39 |
+
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](https://github.com/XLearning-SCU/2021-NeurIPS-NCR) (predecessor of NCL), each image in all datasets were preprocessed using Faster-RCNN detector provided by [Bottom-up Attention Model](https://github.com/peteanderson80/bottom-up-attention) 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](https://openaccess.thecvf.com/content/CVPR2023/papers/Gu_Mobile_User_Interface_Element_Detection_via_Adaptively_Prompt_Tuning_CVPR_2023_paper.pdf) 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%**.
|
40 |
|
41 |
<div align=center>
|
42 |
|