Clarify a term, inspect evidence, or place an idea in context.
inAiLoading page content...
Profile · vision and data
Plan for this page
Li led the creation of ImageNet with collaborators, helping make large, shared datasets and benchmarks central to progress in computer vision.
Contributions
Led the ImageNet project and challenge with a large team, creating a shared image dataset and evaluation setting at unprecedented scale.
Advanced object recognition, scene understanding, visual learning and later work in robotics, healthcare and spatial intelligence.
Co-founded Stanford's Institute for Human-Centered Artificial Intelligence and AI4ALL.
Participated in later work to study and reduce privacy and representation problems in ImageNet.
ImageNet organized millions of web images through a hierarchy derived from WordNet. Its scale and annual challenge helped researchers compare systems and contributed to the 2012 breakthrough of deep convolutional networks in image classification.
The project also shows that datasets are designed artifacts. Category choices, web collection, labor, labels, consent and benchmarks carry assumptions. Subsequent removals, balancing work and face obfuscation are part of the history, not footnotes outside it.
Context, limits, and debates
Sources and scope
Current official profile documenting Li's research, ImageNet leadership, Stanford HAI role and World Labs role; statements such as 'inventor' are institutional attribution and should not erase the project's collaborators.
Primary 2009 paper by Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kai Li and Li Fei-Fei describing the dataset's hierarchy, scale and construction.
Official description of the institute and Li's founding-director role; its human-centered mission is a normative institutional program, not proof that deployed AI is human-centered.
Primary study of privacy risks and face obfuscation in ImageNet, useful for understanding later dataset remediation rather than treating the original resource as fixed or unproblematic.
