Notes on Flickr's dataset

1. flickr8k Image annotation dataset ：
   1. Data set containing 8,000 Zhang image , Each image is paired with five different titles , These titles provide content descriptions of objects and events in the pictures
   2. This data set seems to be related to the image description task （ Generate a text description for the image ） of .
   3. Image subtitles generate excellent data sets that can be used ：flickr8k Data sets , Realistic and relatively small .
2. Flickr30K It's from Flickr The sorted out content downloaded from 30k Pictures and data sets corresponding to description sentences
3. IGEODATA Data sets ：
   1. The data set consists of ten bzip2 Compressed files （yfcc100m_dataset-0.bz2 To yfcc100m_dataset-9.bz2） form , Each file contains 10M That's ok , Each row contains the following tab delimited fields ：* Photo / Video identifier 、* user NSID、* The user nickname 、* Date of shooting 、* Upload date 、* Capture devices 、* title 、* describe 、* user tags（ Comma separated ）、* machine tags（ Comma separated ）、* longitude 、* latitude 、* accuracy 、* Photo / Video page URL、* Photo / Video downloading URL、* License name 、* license URL、* Photo / Video server identifier 、* Photo / Video field identifier 、* Photo / Video confidentiality 、* Photo / Original confidential video 、* Expansion of the original photo 、* Photo / Video Tags （0= Photo ,1= video ）
   2. The field containing free-form text has been URL code . Not all fields have values , Especially the camera 、 title 、 describe 、 Mark 、EXIF、 longitude 、 The latitude and precision fields may be empty . Please note that , The original extension is only meaningful for photos , It doesn't make sense for video （ Please check the first few bytes of the video to determine its file format ）.
   3. In addition to dataset files , Also provided is a photo containing / Video identifier and its corresponding MD5 Hash （yfcc100m_hash.bz2） The file of . These hashes will be used for externally hosted expansion packs （ For example, function 、 notes ）, As an indirect layer , To Hide Photos / Direct access to video information .