https://github.com/li-xirong/flickr8kcn

A bilingual dataset for image captioning
https://github.com/li-xirong/flickr8kcn

Last synced: 3 months ago
JSON representation

A bilingual dataset for image captioning

• Host: GitHub
• URL: https://github.com/li-xirong/flickr8kcn
• Owner: li-xirong
• Created: 2018-05-09T15:31:39.000Z (over 6 years ago)
• Default Branch: master
• Last Pushed: 2020-10-28T13:52:03.000Z (about 4 years ago)
• Last Synced: 2024-08-01T02:24:40.003Z (6 months ago)
• Homepage:
• Size: 5.03 MB
• Stars: 17
• Watchers: 2
• Forks: 11
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

• awesome-self-supervised-multimodal-learning - Link - xirong/flickr8kcn)| (Summary of Common Multimodal Datasets / Image-Text Datasets)

README

        
# Flickr8K-CN

Flickr8K-CN is a bilingual (English-to-Chinese) extension of the popular [Flickr8K](http://nlp.cs.illinois.edu/HockenmaierGroup/Framing_Image_Description/KCCA.html) set, used for evaluating image captioning in a cross-lingual setting.

| Chinese sentences | Flickr8k-train | Flickr8k-val | Flickr8k-test | 

| -----:| -----:| -----:| -----:|

| human written    | :white_check_mark: | :white_check_mark: | :white_check_mark: | 

| human translation     | :x:     |   :x:  | :white_check_mark: | 

| machine translation (baidu)  | :white_check_mark: | :white_check_mark: | :white_check_mark: | 

| machine translation (google) | :white_check_mark: | :white_check_mark: | :white_check_mark: | 

## Data

### Sentences

1. [Original English sentences](data/flickr8kenc.caption.txt)

2. [Chinese sentences written by native Chinese speakers](data/flickr8kzhc.caption.txt)

3. Chinese sentences generated by Baidu translation ([icmr2016 version](data/flickr8kzhb.caption.txt), [version 20160815](data/flickr8kzhb.caption.txt.v20160815))

4. Chinese sentences generated by Google translation ([icmr2016 version](data/flickr8kzhg.caption.txt), [version 20160816](data/flickr8kzhg.caption.txt.v20160816))

5. [Chinese sentences generated by human translation](data/flickr8kzhmtest.captions.txt) (only the test set is covered)

### Dataset split

* imageids of [6K training images](data/flickr8ktrain.txt), [1k validation images](data/flickr8kval.txt), [1k test images](data/flickr8ktest.txt)

### Image features

1. [1,024-dim GoogleNet pool5](http://lixirong.net/data/icmr2016/flickr8k-pygooglenet-pool5_7x7_s1.tar.gz), read by [bigfile.py](https://github.com/li-xirong/jingwei/blob/master/util/simpleknn/bigfile.py)

## Citations

1.  Xirong Li, Weiyu Lan, Jianfeng Dong, Hailong Liu, [Adding Chinese Captions to Images](icmr2016_chisent.pdf), ACM ICMR 2016