Deep Learning for Emotion Recognition on Small Datasets using Transfer Learning

By Hongwei Ng, Viet Dung Nguyen, Vassilios Vonikakis, Stefan Winkler. 2015

634 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, Experimental and Cognitive Psychology

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

In Experimental and Cognitive Psychology

Citations in the corpus

ScaleNet: An Unsupervised Representation Learning Method for Limited Information (2023, methodology): “001] for the ScaleNet to avoid over-fitting [36].”
Generalisation in humans and deep neural networks (2018, background): “1 Secondly, features learned on one task often transfer to only loosely related tasks, such as from classification to saliency prediction [33], emotion recognition [34], medical imaging [35] and a large number of other transfer learning tasks [36].”
Data Augmentation in Emotion Classification Using Generative Adversarial Networks (2017, background): “On the other hand, some classes of emotion images (such as disgust ) have few samples in the training set from the real world than other classes, also some images can be very nuanced, making it diﬃcult even for humans to agree on their correct labeling [22].”
From Facial Expression Recognition to Interpersonal Relation Prediction (2016, methodology): “DCN of [54] - AlexNet [37] pretrained with ImageNet [65] and FER [20] datasets, and finetune on the SFEW training dataset.”
A Brief Review of Facial Emotion Recognition Based on Visual Information (2018, background): “Therefore, this sub-section brieﬂy introduces some popular databases related to FER consisting of 2D and 3D video sequences and still images:”
Automatic Analysis of Facial Actions: A Survey (2019, background): “An alternative to dealing with a low quantity of labelled examples is the use of transfer learning techniques [122].”
Deep Learning vs. Traditional Computer Vision (2019, methodology): “The frame reduction techniques discussed previously such as using SIFT features [26, 32] or optical flow for moving objects [27] to first identify a region of interest are useful with respect to image resolution and also with respect to reducing the time and data required for training.”
Deep Affect Prediction in-the-Wild: Aff-Wild Database and Challenge, Deep Architectures, and Beyond (2018, methodology): “For the second approach we used transfer learning [38], especially of the convolutional and pooling part of the pre-trained networks.”
FaceNet2ExpNet: Regularizing a Deep Face Recognition Net for Expression Recognition (2016, methodology): “Compared with networks adopted in previous works [9], [23], [25], AlexNet [47] or VGG-M [48], the size of our network is fairly small.”
Privileged Information for Modeling Affect In The Wild (2021, background): “At this point, we should clarify that transferring knowledge using LUPI is different from the transfer of learning techniques used in deep learning [23].”
Deep Facial Expression Recognition: A Survey (2018, methodology): “Instead of directly using the pretrained or fine-tuned models to extract features on the target dataset, a multistage fine-tuning strategy [67] can achieve better performance; after the first-stage fine-tuning using FER2013 on pretrained”
Deep learning-based facial emotion recognition for human–computer interaction applications (2021, methodology): “learning is one of the methods in deep learning; in this, a model trained for a particular task is reused for another task by the transfer of knowledge [23].”
Towards Efficient and Privacy-Preserving Federated Deep Learning (2019, methodology): “Deep neural network has been applied in various application scenarios such as computer vision [1], emotion analysis [2] and natural language processing [3] due to its superior accuracy of prediction for training enormous amount of data.”
Reliable Crowdsourcing and Deep Locality-Preserving Learning for Expression Recognition in the Wild (2017, background): “In EmotiW 2015 [8], more competitors have tried deep learning meth-ods: transfer learning was used to solve the problem of small database in [32], hierarchical committee of multi-column DCNNs in [20] gained the best result on SFEW database, LBP features combined with DCNNs structure were proposed…”
Reliable Crowdsourcing and Deep Locality-Preserving Learning for Unconstrained Facial Expression Recognition (2019, methodology): ““SFEW best [33]”, “SFEW second [77]” and “SFEW third [54]” indicate the best single model result of the winner, the runner-up and the second runner-up in EmotiW 2015, respectively.”