Flickr Image Dataset

This Flickr data set is collected from one of the world’s largest image sharing website Filckr. It contains 338389 images uploaded by 3,668 users on and before 2013. Every image is associated with its ID, owner, title, date, tags, description, comments, etc (if exists). Every user is associated with its ID, alias, location, gender, marital status, occupation, etc (if exists).

We employ a method first used in [1] and then adopted in [2]-[5] to extract visual features. Visual features include five dominant color of the image, saturation and its contrast, brightness and its contrast, warm or cool color, clear or dull color, area difference and color difference between foreground and background, texture complexity of foreground and background, etc.

We then employ method used in [1]-[5] to label the emotion of every image automatically according to Ekman’s [6] six basic emotion categories: happiness, surprise, anger, disgust, fear and sadness.

In this way, the data set are summarized into four files: 1) image and its associated information; 2) image visual features; 3) image emotions; 4) users. Details can be seen in related references.

[1] Xiaohui Wang, Jia Jia, Jiaming Yin, Lianhong Cai. Interpretable Aesthetic Features for Affective Image Classification. IEEE International Conference on Image Processing (ICIP), Melbourne, Australia, Sep. 15-18, 2013, pp. 3230-3234.

[2] Boya Wu, Jia Jia, Yang Yang, Peijun Zhao, Jie Tang. Understanding The Emotions Behind Social Images: Inferring With User Demographics. In Proceedings of the 16th International Conference on Multimedia & Expo (ICME'15)

[3] Yang Yang, Jia Jia, Shumei Zhang, Boya Wu, Juanzi Li, and Jie Tang. How Do Your Friends on Social Media Disclose Your Emotions? AAAI, 2014

[4] Yang Yang, Jia Jia, Boya Wu, Jie Tang. Social Role-Aware Emotion Contagion in Image Social Networks. In Proceedings of the thirtieth AAAI Conference on Artificial Intelligence (AAAI'16)

[5] Xiaohui Wang, Jia Jia, Jie Tang, Boya Wu, Lianhong Cai, Lexing Xie. Modeling Emotion Influence in Image Social Networks. IEEE Transactions on Affective Computing, v 6, n 3, p 286-297, July 1, 2015

[6] P.Ekman, “Anargumentforbasicemotions,” Cognition and Emotion, vol. 6, no. 3-4, pp. 169–200, 1992.


In this version we release the original images of the dataset. Images whose url are invalid at time we release the dataset are removed from the v1.0 version.

Database v1.1
Image (Readme)
(216 M)
Image Feature (Readme)
(30.9 M)
Image Word Score (Readme)
(1.49 M)
User (Readme)
(52.4 K)
Whole Dataset
(249 M)

Raw images and urls for the dataset:

Database v1.1
(80.6 G)
Image urls
(3.17 M)


Database v1.0
Image (Readme)
(222 M)
Image Feature (Readme)
(31.7 M)
Image Word Score (Readme)
(1.54 M)
User (Readme)
(52.6 K)
Whole Dataset
(255 M)