InclusiveFaceNet: Improving Face Attribute Detection with Race and Gender Diversity

By Hee Jung Ryu, Hartwig Adam, Margaret Mitchell. 2017

116 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, Safety, Robustness & Interpretability

How researchers see this paper

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

In Computer Vision and Pattern Recognition

In Safety, Robustness & Interpretability

Citations in the corpus

FairFace: Face Attribute Dataset for Balanced Race, Gender, and Age (2019, background): “Studies in algorithmic fairness have focused on either 1) discovering (auditing) existing bias in datasets or systems [50, 7, 30], 2) making a better dataset [39, 1], or 3) designing a better algorithm or model [12, 1, 47, 71, 70].”
Women also Snowboard: Overcoming Bias in Captioning Models (2018, background): “s, we consider the quotient between predicted probabilities for man and gender words (FM is of the same form): FW ( ~w t;I) = P g m2G p( ~w t = g mjw 0:t 1;I) ( P g w2G w p( ~w t = g wjw 0:t 1;I))+ (5) Women also Snowboard: Overcoming Bias in Captioning Models 7 where is a small epsilon value added”
A Step Toward More Inclusive People Annotations for Fairness (2021, background): “, Celeb-A [18], FairFace [14], UTKFace [32]) but may also be useful for studying bias [1, 23].”
Towards fairer datasets: filtering and balancing the distribution of the people subtree in the ImageNet hierarchy (2019, methodology): “Rather than addressing the full complexity of this question, we follow prior work [10, 11, 41, 65, 83] and use a set of discrete categories: Male , Female , and Unsure , in which Unsure is used to both handle ambiguous visual cues as well as to include people with diverse gender expression.”
Examining Gender and Race Bias in Two Hundred Sentiment Analysis Systems (2018, background): “Approaches on how to mitigate inappropriate biases (Schmidt, 2015; Bolukbasi et al., 2016; Kilbertus et al., 2017; Ryu et al., 2017; Speer, 2017; Zhang et al., 2018; Zhao et al., 2018) are also beyond the scope of this paper.”
Diversity in Faces (2019, methodology): “ender probability when gender evidence is occluded in a scene, and otherwise predicts gender from relevant information when present. The problem of gender-neutral smile classication was addressed in [42]. Bias in face detection for skin tone, pose and expression was studied in [52]. Buolamwini et al”
50 Years of Test (Un)fairness: Lessons for Machine Learning (2018, methodology): “Recent approaches have used Fitzpatrick skin type or unsupervised clustering to avoid racial categorizations [7, 55].”
Algorithmic Fairness (2020, background): “However, the task becomes much more complex when a fair description of images is required, such as in multi-label classification tasks [82, 129, 134, 148], in face attribute detection [81, 121], or in the task of image captioning [67].”
Inclusive GAN: Improving Data and Minority Coverage in Generative Models (2020, background): “The biases are rooted in the inevitable imbalance in the dataset [42], which are preserved or even exacerbated by the generative models [57].”
A Framework for Understanding Unintended Consequences of Machine Learning (2019, methodology): “Since this study, other algorithms have been benchmarked on more balanced face datasets, changing the development process to encourage models that perform well across groups (Ryu, Adam, and Mitchell, 2018).”
Towards Fairness in Visual Recognition: Effective Strategies for Bias Mitigation (2019, methodology): “n provide a thorough comparison of existing methods for bias mitigation, including domain adversarial training [46, 41, 1], Reducing Bias Ampliﬁcation [52], and domain conditional training similar to [40]. To the best of our knowledge, no such comparison exists currently as these methods have been e”
Are My Deep Learning Systems Fair? An Empirical Study of Fixed-Seed Training (2021, background): “The fairness of ML and DL systems have gained attention and several studies pointed out fairness issues in datasets (e.g., CelebA [81], GitHub [41], and StackOver-ﬂow [61]) and major ML systems such as language models [13, 92, 95], face recognition [89], and sentiment analysis [88].”
Fair Attribute Classification through Latent Space De-biasing (2020, background): “Some work propose to mitigate bias through adversarial learning [40, 45] or through learning separate classifiers for each protected attribute [33, 41].”
Fair Generative Modeling via Weak Supervision (2019, background): “[38] considers transfer learning of race and gender identities as a form of weak supervision for predicting other attributes on datasets of faces.”
Towards causal benchmarking of bias in face analysis algorithms (2020, background): “The machine learning community is active in analyzing biases of learning models, and how one may train models where bias is mitigated [3, 14, 18, 31, 33, 41, 46, 51, 68], usually by ensuring that performance is equal across certain subgroups of a dataset.”
Understanding Unequal Gender Classification Accuracy from Face Images (2018, background): “otential of decoupled classiﬁcation on demographic groups, which along with task transfer learning has been shown to mitigate biases in classiﬁcation of other facial attributes across race and gender [30]. Finally, the perspectives presented here are limited to the problem of binary gender classiﬁca”
A Review on Fairness in Machine Learning (2022, background): “However, the task becomes much more complex when a fair description of images is required, such as in multi-label classification tasks [117, 193, 202, 221], in face attribute detection [116, 177], or in the task of image captioning [96].”
Deep Learning for Face Recognition: Pride or Prejudiced? (2019, background): “Ryu et al. [32] proposed using transfer learning to develop attribute prediction models inclusive of different race and gender subgroups.”
FairFace: Face Attribute Dataset for Balanced Race, Gender, and Age for Bias Measurement and Mitigation (2021, background): “…have focused on either 1) discovering (auditing) existing bias in datasets or systems [52, 7, 30, 39, 22], 2) making a better dataset [40, 1], or 3) designing a better algorithm or model [12, 1, 49, 72, 71, 62, 41, 27], typically by learning representations invariant to sensitive attributes.”