The third Emotion Recognition in the Wild (EmotiW) 2015 challenge consists of emotion classification challenges, which mimics real-world conditions. There are two sub-challenges:

1. Audio-Video based emotion recognition challenge based on AFEW database [1]
2. Image based static facial expression recognition challenge based on SFEW database [2]

The emotion recognition challenge contains audio-video short clips labelled using a semi-automatic approach defined in [1]. This challenge is continuation from the challenge in EmotiW 2013 & 2014. The task is to assign a single emotion label to the video clip from the six universal emotions (Anger, Disgust, Fear, Happiness, Sad & Surprise) and Neutral. The second challenge contains frame level facial expression labels. The task is to assign a facial expression label to a frame (containing a face) out of the six universal (Anger, Disgust, Fear, Happiness, Sad & Surprise) and Neutral.

The databases (AFEW 5.0 and SFEW 2.0) will be divided into three sets for the challenge: training, validation and testing. For emotion recognition (Challenge 1), there are no separate video-only, audio-only, or audio-video challenges. Participants are free to use either modality or both. Results for all methods will be combined into one set in the end. Participants are allowed to use their own features and classification methods. The labels of the testing set are unknown. For both the challenges, participants will need to adhere to the definition of training, validation and testing sets. In their papers, they may report on results obtained on the training and validation sets, but only the results on the testing set will be taken into account for the overall Grand Challenge results. A paper describing the method is required to be submitted by the participating teams.
For any queries please email at: EmotiW2014@gmail.com

Baseline
1. Audio-Video based emotion recognition challenge
For video, the face is localized using Mixture of Parts framework of Zhu and Ramana 2012 and tracking is performed using IntraFace library. The fiducial points generated by IntraFace are used for aligning the face. Post aligning LBP-TOP features are extracted from non-overlapping spatial 4x4 blocks. The LBP-TOP feature from each block are concatenated to create one feature vector. Chi square kernel based SVM is learnt for emotion classification. The video only baseline system achieves 36.08%. For the Test set the accuracy is 39.33% classification accuracy.

2. Image based static facial expression recognition
Face and facial parts detection is performed using Mixture of Pictorial Structures. Pyramid of Histogram of Gradients and Local Phase Quantisation features are computed on aligned faces. SVM is trained on the vector computed from feature fusion. The classification accuracy is 23.16%35.96%. For the test set, the classification accuracy is 39.13%.

References
1. A. Dhall, R. Goecke, S. Lucey and T. Gedeon, Collecting large, richly annotated facial-expression databases from movies, IEEE Multimedia 2012. [PDF]

2. A. Dhall, R. Goecke, S. Lucey and T. Gedeon, Static Facial Expression in Tough Conditions: Data, Evaluation Protocol and Benchmark, IEEE ICCV BEFIT Workshop 2011. [PDF]


EmotiW 2014 baseline paper -
A. Dhall, R. Goecke, J. Joshi, K. Sikka and T. Gedeon, Emotion Recognition In The Wild Challenge 2014: Baseline, Data and Protocol, ACM ICMI 2014. [PDF]

EmotiW 2013 baseline paper -
A. Dhall, R. Goecke, J. Joshi, M. Wagner and T. Gedeon, Emotion Recognition In The Wild Challenge 2013, ACM ICMI 2013. [PDF]

EmotiW 2014 winner -
M. Liu, R. Wang, S. Li, S. Shan, Z. Huang and X. Chen, Combining Multiple Kernel Methods on Riemannian Manifold for Emotion Recognition in the Wild, ICMI 2014. [PDF]

EmotiW 2013 winner -
S. Ebrahimi Kahou, C. Pal, X. Bouthillier, P. Froumenty, S. Jean, K. Reddy Konda, P. Vincent, A. Courville and Y. Bengio, Combining Modality Specific Deep Neural Network Models for Emotion Recognition in Video, ACM ICMI 2013. [PDF]