Large image datasets: A pyrrhic win for computer vision?

Abstract

In this paper we investigate problematic practices and consequences of large scale vision datasets. We examine broad issues such as the question of consent and justice as well as specific concerns such as the inclusion of verifiably pornographic images in datasets. Taking the ImageNet-ILSVRC-2012 dataset as an example, we perform a cross-sectional model-based quantitative census covering factors such as age, gender, NSFW content scoring, class-wise accuracy, human-cardinality-analysis, and the semanticity of the image class information in order to statistically investigate the extent and subtleties of ethical transgressions. We then use the census to help hand-curate a look-up-table of images in the ImageNet-ILSVRC-2012 dataset that fall into the categories of verifiably pornographic: shot in a non-consensual setting (up-skirt), beach voyeuristic, and exposed private parts. We survey the landscape of harm and threats both society broadly and individuals face due to uncritical and ill-considered dataset curation practices. We then propose possible courses of correction and critique the pros and cons of these. We have duly open-sourced all of the code and the census meta-datasets generated in this endeavor for the computer vision community to build on. By unveiling the severity of the threats, our hope is to motivate the constitution of mandatory Institutional Review Boards (IRB) for large scale dataset curation processes.

Github: https://github.com/vinayprabhu/Dataset_audits. Update on July 23rd: (1) Added in the supplementary section (2) The curators of the Tiny Images dataset decided to withdraw the dataset in response to the previous version of this paper, a change that has duly been reflected in this version. Their statement: https://groups.csail.mit.edu/vision/TinyImages/

References in corpus (21)

• Distilling the Knowledge in a Neural Network
• UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction
• Large Scale GAN Training for High Fidelity Natural Image Synthesis
• Progressive Growing of GANs for Improved Quality, Stability, and Variation
• The History Began from AlexNet: A Comprehensive Survey on Deep Learning Approaches
• The Open Images Dataset V4: Unified image classification, object detection, and visual relationship detection at scale
• RetinaFace: Single-stage Dense Face Localisation in the Wild
• Do ImageNet Classifiers Generalize to ImageNet?
• Learning Robust Global Representations by Penalizing Local Predictive Power
• Lipstick on a Pig: Debiasing Methods Cover up Systematic Gender Biases in Word Embeddings But do not Remove Them
• No Classification without Representation: Assessing Geodiversity Issues in Open Data Sets for the Developing World
• Predictive Inequity in Object Detection
• Data Augmentation Using GANs
• Measuring the Algorithmic Efficiency of Neural Networks
• Tencent ML-Images: A Large-Scale Multi-Label Image Database for Visual Representation Learning
• Are we done with ImageNet?
• Assessing Social and Intersectional Biases in Contextualized Word Representations
• Algorithmic Injustices: Towards a Relational Ethics
• Auditing ImageNet: Towards a Model-driven Framework for Annotating Demographic Attributes of Large-Scale Image Datasets
• Aesthetics of Neural Network Art
• Fonts-2-Handwriting: A Seed-Augment-Train framework for universal digit classification

Cited by in corpus (15)

• The Pile: An 800GB Dataset of Diverse Text for Language Modeling
• Towards decolonising computational sciences
• Differentially Private Learning Needs Better Features (or Much More Data)
• Data and its (dis)contents: A survey of dataset development and use in machine learning research
• Towards Accountability for Machine Learning Datasets: Practices from Software Engineering and Infrastructure
• Unlearnable Examples: Making Personal Data Unexploitable
• Pitfalls in Machine Learning Research: Reexamining the Development Cycle
• The Limits of Global Inclusion in AI Development
• Mitigating Dataset Harms Requires Stewardship: Lessons from 1000 Papers
• An Ethical Highlighter for People-Centric Dataset Creation
• Retiring Adult: New Datasets for Fair Machine Learning
• Insights into Data through Model Behaviour: An Explainability-driven Strategy for Data Auditing for Responsible Computer Vision Applications
• Out-distribution aware Self-training in an Open World Setting
• MONCAE: Multi-Objective Neuroevolution of Convolutional Autoencoders
• The Slodderwetenschap (Sloppy Science) of Stochastic Parrots -- A Plea for Science to NOT take the Route Advocated by Gebru and Bender