Efficient Group-n Encoding and Decoding for Facial Age Estimation

Zichang Tan¹ Jun Wan¹ Zhen Lei¹ Ruicong Zhi² Guodong Guo³ Stan Z. Li¹

¹Institute of Automation, Chinese Academy of Sciences

²School of Computer and Communication Engineering, University of Science and Technology Beijing

³Department of Computer Science and Electrical Engineering, West Virginia University

Different ages are closely related especially among the adjacent ages because aging is a slow and extremely non-stationary process with much randomness. To explore the relationship between the real age and its adjacent ages, an age group-n encoding (AGEn) method is proposed in this paper, where adjacent ages are grouped into the same group and each age corresponds to n groups. The ages grouped into the same group would be regarded as an independent class in the training stage. On this basis, the proposed method transforms age estimation problem into a series of binary classification sub-problems. And a deep Convolutional Neural Networks (CNN) with multiple classifiers is designed to cope with such sub-problems. Later, a Local Age Decoding (LAD) strategy is further presented to accelerate the prediction process, which locally decodes the estimated age value from ordinal classifiers. Besides, to alleviate the imbalance data learning problem of each classifier, a penalty factor is inserted into the unified objective function to favor the minority class. To compare with state-of-the-art methods, we evaluate our proposed method on FG-NET, MORPH II, CACD and Chalearn LAP 2015 Databases and it achieves the best performance.

Updates:

l [new] The pretrained caffemodels on IMDB-WIKI dataset and the trained caffemodel on Chalearn LAP 2016 dataset released.

l [new] The evaluation codes released.

Downloads:

[Code]

1. [Caffemodels]: The caffemodels can be downloaded from this link. (Password: srdj)

2. [Evaluation codes]: The python codes for prediction on Chalearn LAP 2016.

Results:

1. Some prediction samples on Chalearn LAP, FG-NET, Morph and CACD databases.

2. Cumulative Score (CS) results on FG-NET, Morph and CACD databases.

Citation:

Please add a reference if you are using the codes or the pretrained models.

@article{Tan-PAMI-2017,
  author = {Tan, Zichang and Wan, Jun and Lei, Zhen and Zhi, Ruicong and Guo, Guodong and Li, Stan Z.},
  title = {Efficient Group-n Encoding and Decoding for Facial Age Estimation},
  journal = {IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI), 2017},
  year = {2017},
}