Learning from Large-scale Noisy Web Data with Ubiquitous Reweighting for Image Classification
This addresses the costly and time-consuming issue of constructing clean datasets for image classification by enabling effective learning from noisy web data, though it is incremental as it builds on existing reweighting techniques.
The paper tackles the problem of learning image classification from large-scale noisy web data by proposing a Ubiquitous Reweighting Network (URNet) that reweights training instances to mitigate data bias and noise, achieving state-of-the-art performance and ranking first in the WebVision 2018 challenge with 16 million noisy images across 5000 classes.
Many advances of deep learning techniques originate from the efforts of addressing the image classification task on large-scale datasets. However, the construction of such clean datasets is costly and time-consuming since the Internet is overwhelmed by noisy images with inadequate and inaccurate tags. In this paper, we propose a Ubiquitous Reweighting Network (URNet) that learns an image classification model from large-scale noisy data. By observing the web data, we find that there are five key challenges, i.e., imbalanced class sizes, high intra-classes diversity and inter-class similarity, imprecise instances, insufficient representative instances, and ambiguous class labels. To alleviate these challenges, we assume that every training instance has the potential to contribute positively by alleviating the data bias and noise via reweighting the influence of each instance according to different class sizes, large instance clusters, its confidence, small instance bags and the labels. In this manner, the influence of bias and noise in the web data can be gradually alleviated, leading to the steadily improving performance of URNet. Experimental results in the WebVision 2018 challenge with 16 million noisy training images from 5000 classes show that our approach outperforms state-of-the-art models and ranks the first place in the image classification task.