Challenges in Representation Learning: A Report on Three Machine Learning Contests

Goodfellow, Ian J.; Erhan, Dumitru; Carrier, Pierre; Courville, Aaron; Mirza, Mehdi; Hamner, Ben; Cukierski, Will; Tang, Yichuan; Thaler, David S.; Lee, Dong-Hyun; Zhou, Yingbo; Ramaiah, Chetan; Feng, Fan; Li, Ruifan; Wang, Xiaojie; Athanasakis, Dimitris; Shawe‐Taylor, John; Milakov, Maxim; Park, John; Ionescu, Radu Tudor; Popescu, Marius; Grozea, Cristian; Bergstra, James; Xie, Jingjing; Romaszko, Łukasz; Xu, Bing; Zhang, Chuang; Bengio, Yoshua

doi:10.1007/978-3-642-42051-1_16

Cited by 955 publications

(346 citation statements)

References 9 publications

Supporting

Mentioning

340

Contrasting

Unclassified

Order By: Relevance

“…We selected the first 1500 images from FER2013 [36] as our training dataset. The reason for intentional lowered size of training dataset is to enforce the similar model complexity between a source BNN and a target BNN.…”

Section: Training Datasetmentioning

confidence: 99%

Uncertainty Flow Facilitates Zero-Shot Multi-Label Learning in Affective Facial Analysis

2018

View full text Add to dashboard Cite

Featured Application: The proposed Uncertainty Flow framework may benefit the facial analysis with its promised elevation in discriminability in multi-label affective classification tasks. Moreover, this framework also allows the efficient model training and between tasks knowledge transfer. The applications that rely heavily on continuous prediction on emotional valance, e.g., to monitor prisoners' emotional stability in jail, can be directly benefited from our framework. Abstract:To lower the single-label dependency on affective facial analysis, it urges the fruition of multi-label affective learning. The impediment to practical implementation of existing multi-label algorithms pertains to scarcity of scalable multi-label training datasets. To resolve this, an inductive transfer learning based framework, i.e., Uncertainty Flow, is put forward in this research to allow knowledge transfer from a single labelled emotion recognition task to a multi-label affective recognition task. I.e., the model uncertainty-which can be quantified in Uncertainty Flow-is distilled from a single-label learning task. The distilled model uncertainty ensures the later efficient zero-shot multi-label affective learning. On the theoretical perspective, within our proposed Uncertainty Flow framework, the feasibility of applying weakly informative priors, e.g., uniform and Cauchy prior, is fully explored in this research. More importantly, based on the derived weight uncertainty, three sets of prediction related uncertainty indexes, i.e., soft-max uncertainty, pure uncertainty and uncertainty plus are proposed to produce reliable and accurate multi-label predictions. Validated on our manual annotated evaluation dataset, i.e., the multi-label annotated FER2013, our proposed Uncertainty Flow in multi-label facial expression analysis exhibited superiority to conventional multi-label learning algorithms and multi-label compatible neural networks. The success of our proposed Uncertainty Flow provides a glimpse of future in continuous, uncertain, and multi-label affective computing.

show abstract

Section: Training Datasetmentioning

confidence: 99%

Uncertainty Flow Facilitates Zero-Shot Multi-Label Learning in Affective Facial Analysis

2018

View full text Add to dashboard Cite

show abstract

“…Other databases containing posed expressions in controlled conditions include the socalled JAFFE [52], MMI [69] and the GEMEP [4,99] 5 . To the best of our knowledge the only benchmarks that contain samples captured "in-the-wild" are the ones that have been used in the EmotiW series of competitions (the benchmarks are the so-called AFEW and the SFEW datasets [17,18]) and the FER-2013 database [27] 6 . The FER-2013 [27] was created using the Google image search engine to search for images of faces that match a set of 184 emotion-related keywords like blissful, enraged, etc.…”

Section: Databases and Benchmarks For Facial Expression Recognitionmentioning

confidence: 99%

“…The Facial Expression Recognition 2013 (FER-2013) database introduced in the Challenges in Representation Learning (ICML 2013) [27]. The dataset was created using the Google image search API targeting images of faces.…”

Section: Introductionmentioning

confidence: 99%

Facial Affect “In-the-Wild”: A Survey and a New Database

Zafeiriou

Papaioannou

Kotsia

et al. 2016

2016 IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)

View full text Add to dashboard Cite

Well-established databases and benchmarks have been developed in the past 20 years for automatic facial behaviour analysis. Nevertheless, for some important problems regarding analysis of facial behaviour, such as (a) estimation of affect in a continuous dimensional space (e.g., valence and arousal) in videos displaying spontaneous facial behaviour and (b) detection of the activated facial muscles (i.e., facial action unit detection), to the best of our knowledge, well-established in-the-wild databases and benchmarks do not exist. That is, the majority of the publicly available corpora for the above tasks contain samples that have been captured in controlled recording conditions and/or captured under a very specific milieu. Arguably, in order to make further progress in automatic understanding of facial behaviour, datasets that have been captured in in-the-wild and in various milieus have to be developed. In this paper, we survey the progress that has been recently made on understanding facial behaviour inthe-wild, namely the datasets and methodologies that have been developed thus far, while paying particular attention to recently proposed deep learning techniques. Finally, we attempt a significant step further by proposing a novel, comprehensive benchmark that can be utilized for evaluating and training various methodologies for the problems of facial affect, behaviour analysis and understanding "inthe-wild". To the best of our knowledge, this is the first benchmark proposed for measuring continuous affect in the valence-arousal space "in-the-wild".

show abstract

“…We have selected Facial Expression Recognition 2013 (FER-2013) dataset [6]. FER-2013 was created by Pierre Luc Carrier.…”

Section: Fer-2013 Datasetmentioning

confidence: 99%

Feature Acquisition From Facial Expression Image Using Convolutional Neural Networks

Nishime

Endo

Yamada

et al. 2016

JRNAL

View full text Add to dashboard Cite

In this study, we carried out the facial expression recognition from facial expression dataset using Convolutional Neural Networks (CNN). In addition, we analyzed intermediate outputs of CNN. As a result, we have obtained a emotion recognition score of about 58%; two emotions (Happiness, Surprise) recognition score was about 70%. We also confirmed that specific unit of intermediate layer have learned the feature about Happiness. This paper details these experiments and investigations regarding the influence of CNN learning from facial expression.

show abstract

Challenges in Representation Learning: A Report on Three Machine Learning Contests

Cited by 955 publications

References 9 publications

Uncertainty Flow Facilitates Zero-Shot Multi-Label Learning in Affective Facial Analysis

Uncertainty Flow Facilitates Zero-Shot Multi-Label Learning in Affective Facial Analysis

Facial Affect “In-the-Wild”: A Survey and a New Database

Feature Acquisition From Facial Expression Image Using Convolutional Neural Networks

Contact Info

Product

Resources

About