Justin Johnson

Career

World Labs lists Johnson among its founders, together with Fei-Fei Li, Ben Mildenhall and Christoph Lassner, describing the founding team as working across machine learning, generative AI, computer vision and graphics. Johnson's GitHub profile names World Labs as Johnson's organization.¹²

Research

Johnson's early papers, written with Fei-Fei Li and others, include DenseCap (2015) on dense image captioning with Andrej Karpathy, the perceptual-losses paper (2016) on real-time style transfer and super-resolution, and the CLEVR dataset (2016) for diagnosing visual reasoning. Johnson's GitHub account hosts the reference code for the perceptual-losses paper in the fast-neural-style repository, alongside the densecap repository.⁵³⁴⁸²

Later work moved toward 3D understanding: Johnson co-authored Mesh R-CNN (2019) with Georgia Gkioxari and Jitendra Malik, which predicts 3D meshes from images, and the 2020 paper describing the PyTorch3D library.⁶⁷

Work and contributions

• Perceptual Losses for Real-Time Style Transfer and Super-Resolution Published research

  First author, with Alexandre Alahi and Fei-Fei Li, of the 2016 paper that trains feed-forward networks for image transformation tasks such as style transfer and super-resolution using perceptual loss functions.³⁸

• CLEVR: A Diagnostic Dataset for Compositional Language and Elementary Visual Reasoning Open data

  First author of the 2016 paper introducing CLEVR, a diagnostic dataset for testing compositional language and elementary visual reasoning.⁴

• Accelerating 3D Deep Learning with PyTorch3D Open-source code

  Co-author of the 2020 paper describing PyTorch3D, a library for 3D deep learning.⁷

Timeline

1. Nov 2015Posted DenseCap, on fully convolutional localization networks for dense captioning, as first author.⁵
2. Mar 2016Posted the perceptual-losses paper on real-time style transfer and super-resolution as first author.³
3. Dec 2016Posted the CLEVR visual reasoning dataset paper as first author.⁴
4. Jun 2019Co-authored Mesh R-CNN with Georgia Gkioxari and Jitendra Malik.⁶
5. Jul 2020Co-authored the paper describing the PyTorch3D library for 3D deep learning.⁷

Where they work

• World Labs 4 people in the wiki

Related people

• Ben MildenhallAlso at World Labs
• Christoph LassnerAlso at World Labs
• Fei-Fei LiAlso at World Labs
• Luke Melas-KyriaziComputer vision · Image, video, and 3D generation
• Miika AittalaImage, video, and 3D generation · Computer vision
• Patrick EsserImage, video, and 3D generation · Computer vision
• Angjoo KanazawaComputer vision · Image, video, and 3D generation
• Phillip IsolaComputer vision · Image, video, and 3D generation
• Ron MokadyImage, video, and 3D generation · Computer vision
• David LuebkeImage, video, and 3D generation · Computer vision