Deep Label Distribution Learning with Label Ambiguity

6 Nov 2016  ·  Bin-Bin Gao, Chao Xing, Chen-Wei Xie, Jianxin Wu, Xin Geng ·

Convolutional Neural Networks (ConvNets) have achieved excellent recognition performance in various visual recognition tasks. A large labeled training set is one of the most important factors for its success. However, it is difficult to collect sufficient training images with precise labels in some domains such as apparent age estimation, head pose estimation, multi-label classification and semantic segmentation. Fortunately, there is ambiguous information among labels, which makes these tasks different from traditional classification. Based on this observation, we convert the label of each image into a discrete label distribution, and learn the label distribution by minimizing a Kullback-Leibler divergence between the predicted and ground-truth label distributions using deep ConvNets. The proposed DLDL (Deep Label Distribution Learning) method effectively utilizes the label ambiguity in both feature learning and classifier learning, which help prevent the network from over-fitting even when the training set is small. Experimental results show that the proposed approach produces significantly better results than state-of-the-art methods for age estimation and head pose estimation. At the same time, it also improves recognition performance for multi-label classification and semantic segmentation tasks.

PDF Abstract

Results from the Paper

Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Head Pose Estimation AFLW DLDL (KL) MAE 9.78 # 6
Head Pose Estimation BJUT-3D Ours DLDL (KL) MAE 0.09 # 1
Age Estimation ChaLearn 2015 DLDL+VGG-Face MAE 3.51 # 5
e-error 0.31 # 6
Age Estimation MORPH Album2 DLDL+VGG-Face MAE 2.42±0.01 # 5
Age Estimation MORPH Album2 DLDL+VGG-Face (KL, Max)3 MAE 2.42 # 5
Multi-Label Classification PASCAL VOC 2007 Ours PF-DLDL mAP 93.4 # 13
Semantic Segmentation PASCAL VOC 2011 DLDL-8s+CRF Mean IoU 67.6 # 1
Multi-Label Classification PASCAL VOC 2012 Ours PF-DLDL mAP 92.4 # 3
Semantic Segmentation PASCAL VOC 2012 DLDL-8s+CRF Mean IoU 67.1 # 1
Head Pose Estimation Pointing'04 Ours DLDL (KL) MAE 4.64 # 1


No methods listed for this paper. Add relevant methods here