Deep Label Distribution Learning With Label Ambiguity

By Bin-Bin Gao, Chao Xing, Chen-Wei Xie, Jianxin Wu, Xin Geng. IEEE Transactions on Image Processing, 2016

500 citations recorded by the source corpus.

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, multilabel 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 deep label distribution learning (DLDL) method effectively utilizes the label ambiguity in both feature learning and classifier learning, which help prevent the network from overfitting even when the training set is small. Experimental results show that the proposed approach produces significantly better results than the 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.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/02567fd428a675ca91a0c6786f47f3e35881bcbd

How researchers see this paper

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

In Artificial Intelligence

In Computer Vision and Pattern Recognition

Citations in the corpus

PhysFormer: Facial Video-based Physiological Measurement with Temporal Difference Transformer (2021, unknown): “The label distribution loss can be formulated as L LD = KL( p , Softmax( ˆp )) , where divergence measure KL( · ) denotes the Kullback-Leibler (KL) divergence [16], and ˆp is the power spectral density (PSD) of predicted rPPG signals.”
Deep learning with noisy labels: exploring techniques and remedies in medical image analysis (2019, methodology): “Label cleaning and preprocessing [70] - image classification most applications, including disease and pathology classification ([71]; experiments in Section V-B) and lesion detection and segmentation (experiments in Section V-A) [72] - image classification [73] - image classification [74] - image cl”
EEG-Based Emotion Recognition Using Regularized Graph Neural Networks (2019, unknown): “Our proposed Emo-tionDL regularizer is inspired by [53], which applies distribution learning to classify ambiguous images.”
Amalgamating Filtered Knowledge: Learning Task-customized Student from Multi-task Teachers (2019, methodology): “…of the proposed method against the following state-of-the-art approaches: HCP-2000 [Wei et al., 2014], HCP-VGG [Wei et al., 2016], RLSD+ft-RPN [Zhang et al., 2018], CNN+RNN [Wang et al., 2016], CNN+LSTM [Wang et al., 2017], PF-DLDL [Gao et al., 2017a] and the method of [Zhou et al., 2018].”
Probabilistic End-To-End Noise Correction for Learning With Noisy Labels (2019, methodology): “• We propose a variant of the DLDL method [6], which is essential for correcting noise contained in our label distributions.”
Robust Lightweight Facial Expression Recognition Network with Label Distribution Training (2021, methodology): “Previous work acquired label distribution by utilizing distribution annotations (Zhou, Xue, and Geng 2015; Jia et al. 2019), face afﬁnity graph (He et al. 2017), label smoothing (Gao et al. 2017; Ling and Geng 2019) or auxiliary label space graphs (Chen et al. 2020a”