Facial Age Estimation by Learning from Label Distributions

By Xin Geng, K. Smith‐Miles, Zhi-Hua Zhou. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2010

626 citations recorded by the source corpus.

One of the main difficulties in facial age estimation is that the learning algorithms cannot expect sufficient and complete training data. Fortunately, the faces at close ages look quite similar since aging is a slow and smooth process. Inspired by this observation, instead of considering each face image as an instance with one label (age), this paper regards each face image as an instance associated with a label distribution. The label distribution covers a certain number of class labels, representing the degree that each label describes the instance. Through this way, one face image can contribute to not only the learning of its chronological age, but also the learning of its adjacent ages. Two algorithms, named IIS-LLD and CPNN, are proposed to learn from such label distributions. Experimental results on two aging face databases show remarkable advantages of the proposed label distribution learning algorithms over the compared single-label learning algorithms, either specially designed for age estimation or for general purpose.

Subfields: Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/0c741fa0966ba3ee4fc326e919bf2f9456d0cd74

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

Citations in the corpus

Deep Label Distribution Learning With Label Ambiguity (2016, background): “Firstly, it is difficult (even for domain experts) to provide exact labels to some tasks.”
Age , Gender and Race Estimation from Unconstrained Face Images (2014, methodology): “16 We perform exact age estimation on the FG-NET database following the widely used leave-one-person-out (LOPO) protocol [2], [3], [24].”
Binary relevance for multi-label learning: an overview (2018, background): “Di ﬀ erent forms of auxiliary information exist, including ordinal scale over each class label [40], full ranking over relevant class labels [41], importance distribution over all class labels [43,44], and oracle feedbacks over queried labels of unlabeled examples [45].”
PhysFormer: Facial Video-based Physiological Measurement with Temporal Difference Transformer (2021, neutral): “Similar to the facial age estimation task [17, 18] that faces at close ages look quite similar, facial rPPG signals with close HR values usually have similar periodicity.”
Heterogeneous Face Attribute Estimation: A Deep Multi-Task Learning Approach (2017, background): “Due to limited space, we refer interested readers to reviews of the prior work on single facial attribute estimation in [10], [12], [14], [15], [39], [40], [41], [42].”
ChaLearn Looking at People and Faces of the World: Face AnalysisWorkshop and Challenge 2016 (2016, methodology): “They fine-tune this CNN on the IMDB-Wiki dataset [40] for the age estimation task using the label distribution encoding [20].”
Beyond Bilinear: Generalized Multimodal Factorized High-Order Pooling for Visual Question Answering (2017, background): “answers to the LDL problem [37] with the answer distribu-”
Demographic Estimation from Face Images: Human vs. Machine Performance (2015, background): “The appearance-based approach, such as in [17], [19], [21], [26], [27], [34], [43], [48], utilizes facial appearance (both texture and shape) to differentiate faces among Fig.”