天天看點
傳回

表情識别資料集整理

原文見http://blog.csdn.net/computerme/article/details/49469767

  1. CK and CK+ 

    It contains 97 subjects, which posed in a lab situation for the six universal expressions and the neutral expression. Its extension CK+ contains 123 subjects but the new videos were shot in a similar environment. 

    Reference: P. Lucey, J. F. Cohn, T. Kanade, J. Saragih, Z. Ambadar, and I. Matthews, “The Extended Cohn-Kanade Dataset (CK+): A complete dataset for action unit and emotion-specified expression,” in Proceedings of IEEE Conference on Computer Vision and Pattern Recognition Workshops CVPR4HB’10, 2010, pp. 94–101. 

    Website: http://www.pitt.edu/~emotion/ck-spread.htm 

    Modalities: Visual 

    說明: ck隻有靜态圖檔，CK+包括視訊。表情标簽為分類離散值。

  2. JAFFE 

    It contains 219 images of 10 Japanese females. However, it has a limited number of samples, subjects and has been created in a lab controlled environment. 

    Website: http://www.kasrl.org/jaffe.html 

    Modalities: visual 

    說明: 隻有219張表情圖檔。表情标簽為分類離散值。

  3. HUMAINE Database 

    Datafiles containing emotion labels, gesture labels, speech labels and FAPS all readable in ANVI（标簽等資訊要用ANVI工具才能打開） 

    Modalities: Audio+visual + gesture 

    Website: http://emotion-research.net/download/pilot-db/ 

    說明: 下載下傳資料集後裡面隻有視訊，沒有标簽等資訊。

  4. Recola database 

    Totally 34 subjects; 14 male, 20 female 

    Reference: FABIEN R., ANDREAS S., JUERGEN S., DENIS L.. Introducing the RECOLA multimodal corpus of collaborative and affective interactions[C]//10th IEEE Int’l conf. and workshops on automatic face and gesture recognition. Shanghai, CN: IEEE Press, 2013:1-8. 

    Website: http://diuf.unifr.ch/diva/recola/index.html 

    Modalities: Audio+visual+ EDA, ECG（生理模态） 

    說明: 資料集共34個視訊，表情标簽為Arousal-Valence的連續值。标簽存在csv檔案裡。

  5. MMI 

    The database consists of over 2900 videos and high-resolution still images of 75 subjects. It is fully annotated for the presence of AUs in videos (event coding), and partially coded on frame-level, indicating for each frame whether an AU is in either the neutral, onset, apex or offset phase. A small part was annotated for audio-visual laughters. The database is freely available to the scientific community. 

    Reference: 

    a) Induced Disgust, Happiness and Surprise: an Addition to the MMI Facial Expression Database 

    M. F. Valstar, M. Pantic. Proceedings of Int’l Conf. Language Resources and Evaluation, Workshop on EMOTION. Malta, pp. 65 - 70, May 2010. 

    b) Web-based database for facial expression analysis,M. Pantic, M. F. Valstar, R. Rademaker, L. Maat. Proceedings of IEEE Int’l Conf. Multimedia and Expo (ICME’05). Amsterdam, The Netherlands, pp. 317 - 321, July 2005. 

    Modalities: visual(視訊) 

    Website: http://mmifacedb.eu/ 

    　　　　  http://ibug.doc.ic.ac.uk/research/mmi-database/ 

    說明: 該資料集很大，全部包括2900個視訊，标簽主要是AU的标簽，标簽存在xml檔案裡。

  6. NVIE 中科大采集的一個資料集 

    中科大NVIE資料集包括自發表情庫和人為表情庫，本實驗采用其中的自發表情庫。自發表情庫是通過特定視訊誘發并在三種光照下（正面、左側、右側光照）采集的表情庫，其中正面光照103人，左側光照99人，右側光照103人。每種光照下，每人有六種表情（喜悅、憤怒、哀 傷、恐懼、厭惡、驚奇）中的三種以上，每種表情的平靜幀以及最大幀都已挑出 

    Reference: WANG Shangfei, LIU Zhilei, LV Siliang, LV Yanpeng, et al. A Natural Visible and Infrared Facial Expression Database for Expression Recognition and Emotion Inference[J]. IEEE Transactions on Multimedia, 2010, 12(7): 682-691. 

    Website: http://nvie.ustc.edu.cn/ 

    Modalities: visual(圖檔) 

    說明: 标簽以Excel檔案給出，标簽包括表情類的強度，如disgust的表情強度。标簽還包括Arousal-Valence标簽。

  7. RU-FACS database 

    This database consists of spontaneous facial expressions from multiple views, with ground truth FACS codes provided by two facial expression experts. 

    We have collected data from 100 subjects, 2.5 minutes each. This database constitutes a significant contribution towards the 400-800 minute database recommended in the feasibility study for fully automating FACS. To date we have human FACS coded the upper faces of 20% the subjects. 

    Reference: M. S. Bartlett, G. Littlewort, M. G. Frank, C. Lainscsek, I. R. Fasel, and J. R. Movellan, “Automatic recognition of facial actions in spontaneous expressions,” Journal of Multimedia, vol. 1, no. 6, pp. 22–35, 2006. 3, 5 

    Website: http://mplab.ucsd.edu/grants/project1/research/rufacs1-dataset.html 

    說明: 該資料集的标簽是FACS編碼的标簽（隻有部分視訊才有标簽），目前該資料集還未向研究者公開。

  8. Belfast naturalistic database 

    The Belfast database consists of a combination of studio recordings and TV programme grabs labelled with particular expressions. The number of TV clips in this database is sparse 

    Modalities: Audio-visual(視訊) 

    Reference: E. Douglas-Cowie, R. Cowie, and M. Schr¨oder, “A New Emotion Database: Considerations, Sources and Scope,” in ISCAITRW on Speech and Emotion, 2000, pp. 39–44. 

    Website: http://sspnet.eu/2010/02/belfast-naturalistic/ 

    說明: 資料集為視訊，視訊包括speech的情感識别

  9. GEMEP Corpus 

    The GEneva Multimodal Emotion Portrayals (GEMEP) is a collection of audio and video recordings featuring 10 actors portraying 18 affective states, with different verbal contents and different modes of expression. 

    Modalities: Audio-visual 

    Reference: T. B¨anziger and K. Scherer, “Introducing the Geneva Multimodal Emotion Portrayal (GEMEP) Corpus,” in Blueprint for affective computing: A sourcebook, K. Scherer, T. B¨anziger, and E. Roesch, Eds. Oxford, England: Oxford University Press, 2010 

    Website: http://www.affective-sciences.org/gemep 

     　　　　 http://sspnet.eu/2011/05/gemep-fera/ 

    說明: FERA2011比賽采用此資料集，标簽主要是分類。

  10. Paleari 

    Reference: M. Paleari, R. Chellali, and B. Huet, “Bimodal emotion recognition,” in Proceeding of the Second International Conference on Social Robotics ICSR’10, 2010, pp. 305–314. 

    該資料集我沒找到它的官網，我檢視了上面那個引用文章的摘要發現那篇文章不是介紹表情資料集的。那個文章在springer上，學校的網隻能查到到摘要和第一章。

  11. VAM corpus 

    The VAM corpus consists of 12 hours of recordings of the German TV talk-show “Vera am Mittag” (Vera at noon). They are segmented into broadcasts, dialogue acts and utterances, respectively. This audio -visual speech corpus contains spontaneous and very emotional speech recorded from unscripted, authentic discussions between the guests of the talk-show 

    Modalities: Audio-visual 

    Reference: M. Grimm, K. Kroschel, and S. Narayanan, “The Vera am Mittag German audio-visual emotional speech database,” in IEEE International Confernce on Multimedia and Expo ICME’08, 2008, pp. 865–868 

    Website: http://emotion-research.net/download/vam 

    說明: 該資料集主要是speech視訊，标簽為連續值，具體包括三個次元：valence (negative vs. positive), activation (calm vs. excited) and dominance (weak vs. strong)。

  12. SSPNet Conflict Corpus（嚴格意義上不是表情識别資料集） 

    The “SSPNet Conflict Corpus” includes 1430 clips (30 seconds each) extracted from 45 political debates televised in Switzerland. The clips are in French 

    Modalities: Audio-visual 

    Reference: S.Kim, M.Filippone, F.Valente and A.Vinciarelli “Predicting the Conflict Level in Television Political Debates: an Approach Based on Crowdsourcing, Nonverbal Communication and Gaussian Processes“ Proceedings of ACM International Conference on Multimedia, pp. 793-796, 2012. 

    Website: http://www.dcs.gla.ac.uk/vincia/?p=270 

    說明: 該資料集主要是政治辯論中的視訊，标簽為conflict level。

  13. Semaine database 

    The database contains approximately 240 character conversations, and recording is still ongoing. Currently approximately 80 conversations have been fully annotated for a number of dimensions in a fully continuous way using FeelTrace. 

    Website: http://semaine-db.eu/ 

    Modalities: Audio-visual 

    Reference: The SEMAINE database: Annotated multimodal records of emotionally coloured conversations between a person and a limited agent G. Mckeown, M. F. Valstar, R. Cowie, M. Pantic, M. Schroeder. IEEE Transactions on Affective Computing. 3: pp. 5 - 17, Issue 1. April 2012. 

    說明: 通過人機對話來觸發的視訊，标簽為連續的情感次元值，不是分類。

  14. AFEW database(Acted Facial Expressions In The Wild) 

    Acted Facial Expressions In The Wild (AFEW) is a dynamic temporal facial expressions data corpus consisting of close to real world environment extracted from movies. 

    Reference: Abhinav Dhall, Roland Goecke, Simon Lucey, Tom Gedeon, Collecting Large, Richly Annotated Facial-Expression Databases from Movies, IEEE Multimedia 2012. 

    Website: https://cs.anu.edu.au/few/AFEW.html 

    Modalities: Audio-visual(電影剪輯片斷) 

    說明: 該資料集的内容為從電影中剪輯的包含表情的視訊片段，表情标簽為六類基本表情+中性表情，annotation的資訊儲存在xml檔案中。 

    AFEW資料集為Emotion Recognition In The Wild Challenge (EmotiW)系列情感識别挑戰賽使用的資料集，該比賽從2013開始每年舉辦一次。 

    EmotiW官網：https://cs.anu.edu.au/few/

  15. SFEW database(Static Facial Expressions in the Wild) 

    Static Facial Expressions in the Wild (SFEW) has been developed by selecting frames from AFEW 

    Reference: Abhinav Dhall, Roland Goecke, Simon Lucey, and Tom Gedeon. Static Facial Expressions in Tough Conditions: Data, Evaluation Protocol And Benchmark, First IEEE International Workshop on Benchmarking Facial Image Analysis Technologies BeFIT, IEEE International Conference on Computer Vision ICCV2011, Barcelona, Spain, 6-13 November 2011 

    Website: https://cs.anu.edu.au/few/AFEW.html 

    Modalities: Visual 

    說明: 該資料集是從AFEW資料集中抽取的有表情的靜态幀，表情标簽為六類基本表情+中性表情，annotation的資訊儲存在xml檔案中。

  16. AVEC系列資料集 

    AVEC是從2011開始每一年舉辦一次的表情識别挑戰賽，表情識别的模型主要采用的連續情感模型。其中AVEC2012使用的情感次元為Arousal、Valence、Expectancy、Power； AVEC2013的情感次元為Valence和Arousal；AVEC2014的情感次元Valence、Arousal和Dominance。 

    AVEC2013和AVEC2014引入了depression recognition. 

    Modalities: Audio-visual 

    Website: 

    http://sspnet.eu/avec2011/ 

    http://sspnet.eu/avec2012/ 

    http://sspnet.eu/avec2013/ 

    http://sspnet.eu/avec2014/ 

    Reference: Michel Valstar , Björn W. Schuller , Jarek Krajewski , Roddy Cowie , Maja Pantic, AVEC 2014: the 4th international audio/visual emotion challenge and workshop, Proceedings of the ACM International Conference on Multimedia, November 03-07, 2014, Orlando, Florida, USA 

    說明:标簽主要是針對的情感次元，通過csv的形式給出的。

  17. LIRIS-ACCEDE資料集 

    LIRIS-ACCEDE資料集主要包含三個部分: 

    Discrete LIRIS-ACCEDE - Induced valence and arousal rankings for 9800 short video excerpts extracted from 160 movies. Estimated affective scores are also available. 

    Continuous LIRIS-ACCEDE - Continuous induced valence and arousal self-assessments for 30 movies. Post-processed GSR measurements are also available. 

    MediaEval 2015 affective impact of movies task - Violence annotations and affective classes for the 9800 excerpts of the discrete LIRIS-ACCEDE part, plus for additional 1100 excerpts used to extend the test set for the MediaEval 2015 affective impact of movies task. 

    Modalities: Audio-visual 

    Website: 

    http://liris-accede.ec-lyon.fr/index.php 

    Reference: 

    Y. Baveye, E. Dellandrea, C. Chamaret, and L. Chen, “LIRIS-ACCEDE: A Video Database for Affective Content Analysis,” in IEEE Transactions on Affective Computing, 2015. 

    Y. Baveye, E. Dellandrea, C. Chamaret, and L. Chen, “Deep Learning vs. Kernel Methods: Performance for Emotion Prediction in Videos,” in 2015 Humaine Association Conference on Affective Computing and Intelligent Interaction (ACII), 2015 

    M. Sjöberg, Y. Baveye, H. Wang, V. L. Quang, B. Ionescu, E. Dellandréa, M. Schedl, C.-H. Demarty, and L. Chen, “The mediaeval 2015 affective impact of movies task,” in MediaEval 2015 Workshop, 2015 

    說明: 該資料集既有離散的情感資料又有基于次元的情感資料。

幾個重點參考的網站

http://emotion-research.net/wiki/Databases 

http://sspnet.eu/category/sspnet_resource_categories/resource_type_classes/dataset/ 

http://ibug.doc.ic.ac.uk/resources 

http://www.ecse.rpi.edu/~cvrl/database/other_facial_expression.htm

DL/深度學習
上一篇: keras實作RNN手寫字識别
下一篇: PaddleOCR使用筆記之快速使用第1步，安裝環境第2步，下載下傳推理模型