Deep Classifiers from Image Tags in the Wild

By Hamid Izadinia, Bryan C. Russell, Ali Farhadi, Matthew D. Hoffman, Aaron Hertzmann. 2015

63 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

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

Exploring the Limits of Weakly Supervised Pretraining (2018, background): “3 What is the effect of hashtag label noise on model accuracy? A major difference between hashtag supervision and the labels provided in datasets such as ImageNet is that hashtag supervision is inherently noisy : users may apply hashtags that are irrelevant to the visual content of the image, or the”
Revisiting Unreasonable Effectiveness of Data in Deep Learning Era (2017, background): “In order to overcome the bottleneck, there have been recent efforts on visual representation learning using web-supervision [2, 5, 6, 9, 21, 23, 24, 27] or unsupervised [10, 11, 31, 32, 34, 42, 43] paradigms.”
Learning Video Representations from Textual Web Supervision (2020, methodology): “Other approaches use user-deﬁned keywords or tags [20, 28, 43, 19] or algorithmically-generated topics [1, 31] to the same end.”
Learning Visual Features from Large Weakly Supervised Data (2015, methodology): “Izadinia et al. [20] also ﬁnetune pre-trained convolutional networks on a dataset of Flickr images using a vocabulary of 5 , 000 words.”
Large-Scale Weakly-Supervised Pre-Training for Video Action Recognition (2019, background): “Dataset sources: Many prior approaches use Internet as a natural source of content [6, 12, 13, 18, 24, 25, 34, 37, 44, 46, 51, 58, 67] and the associated search queries, hashtags, or user-comments as labels to construct datasets.”
Improving Pairwise Ranking for Multi-label Image Classification (2017, methodology): “Multiple approaches proposed CNN-based techniques for multi-label classification [29, 34, 11].”
Deep learning with noisy labels: exploring techniques and remedies in medical image analysis (2019, background): “A large number of studies keep the model architecture, training data, and training procedures largely intact and only change the loss function [69].”
Recognition From Web Data: A Progressive Filtering Approach (2018, unknown): “[19] perform direct learning from image tags in the wild and achieve improved performance.”
Generating Diverse Image Datasets with Limited Labeling (2016, background): “To address the issues stated above and inspired by streams of images available on the web, there has been lot of recent interest in learning directly from noisy web data [20, 28, 18, 4, 14, 15, 5] or automatically curating web images for creating a dataset [33, 23, 31, 6, 1].”
Image Classification with Deep Learning in the Presence of Noisy Labels: A Survey (2019, background): “EM [104, 105, 26], trace regularizer [106], crowd-layer [107], image difficulty estimate [108] consistency with network [109], omitting probability variable [110] softmax layer per labeler [25]”
CU-Net: Component Unmixing Network for Textile Fiber Identification (2019, methodology): “Izadinia et al. (2015) proposed a robust logistic loss function to train CNNs from user-provided tags.”