AI People Wiki

Builders & open source

Francisco Massa

Emeritus TorchVision maintainer and co-author of Meta's DINOv3, DETR and Detectron2

Francisco Massa lists Facebook AI Research as employer on GitHub. PyTorch's persons-of-interest page names Massa as an emeritus maintainer of TorchVision, the PyTorch computer vision library. Massa is a co-author of Meta's DINOv3 vision foundation model (2025), whose code and weights are released under the DINOv3 License, the second author of DETR (End-to-End Object Detection with Transformers), and is listed in the citation entry for the Detectron2 library.123456

At a glance

Current role
None confirmed in the sources
Work listed
5
Sources cited
14
Profile reviewed

Research

Massa's early papers, written with Mathieu Aubry and Renaud Marlet, studied convolutional networks for detecting objects and estimating their pose and viewpoint, comparing approaches in 2014 and designing a multi-task network for viewpoint estimation in 2016.78

Massa was second author of DETR, which treats object detection as set prediction with a transformer, and co-authored DeiT, a method for training image transformers efficiently with distillation. Later work included the self-supervised vision models DINOv2 and DINOv3, a hybrid transformer for music source separation, and the CWM open-weights code model.591031112

Open-source work

Massa is the third author of the 2019 PyTorch paper describing the library's imperative, Pythonic programming style. PyTorch's persons-of-interest page lists Massa as an emeritus maintainer of TorchVision, and Massa is named in the citation entry of the Detectron2 detection and segmentation library.1326

Work and contributions

  • DINOv3 Open weights

    Listed among the authors of the DINOv3 paper (2025) and in the repository README, which credits Meta AI Research, FAIR, and releases the code and model weights under the DINOv3 License.34

  • TorchVision Open-source code

    Listed as an emeritus TorchVision maintainer (GitHub fmassa) on PyTorch's persons-of-interest page.2

  • DETR Open-source code

    Second author of "End-to-End Object Detection with Transformers" (2020); the DETR repository is released under the Apache 2.0 license.514

  • Detectron2 Open-source code

    Listed among the authors in the citation entry of Detectron2, FAIR's Apache-2.0 detection and segmentation library.6

  • PyTorch Open-source code

    Third author of the 2019 paper "PyTorch: An Imperative Style, High-Performance Deep Learning Library", which describes the design and implementation of the library.13

Timeline

  1. Posted a comparative study of convolutional networks for joint object detection and pose estimation as first author.7
  2. Posted "Crafting a multi-task CNN for viewpoint estimation" as first author.8
  3. Co-authored "PyTorch: An Imperative Style, High-Performance Deep Learning Library".13
  4. Co-authored DETR, end-to-end object detection with transformers, as second author.5
  5. Co-authored DeiT, on training data-efficient image transformers with distillation.9
  6. Co-authored "Hybrid Transformers for Music Source Separation".11
  7. Co-authored DINOv2, on learning robust visual features without supervision.10
  8. Co-authored the DINOv3 vision foundation model paper.3
  9. Co-authored CWM, an open-weights language model for research on code generation with world models.12

Documented roles

The sources do not confirm a current role; these are the roles they document.

Companies in their history

  • Meta 106 people in the wiki

Suggest a correction or removal

Profiles cover public professional work only, and each statement cites a source. If something is out of date, missing or wrong, tell us and link a source. People can also ask for their profile to be corrected or removed.

References

14 sources · profile reviewed Oct 21, 2026 · data updated Oct 21, 2026

  1. fmassa (Francisco Massa) on GitHubGitHub · Official page · accessed Oct 21, 2026
  2. PyTorch Governance — Maintainers (persons of interest), TorchVision sectionPyTorch · Documentation · accessed Oct 21, 2026
  3. DINOv3 (arXiv 2508.10104)arXiv · Paper · published Aug 13, 2025 · accessed Oct 21, 2026
  4. facebookresearch/dinov3 README (authors and DINOv3 License)Meta (GitHub repository) · Repository · accessed Oct 21, 2026
  5. End-to-End Object Detection with Transformers (arXiv 2005.12872)arXiv · Paper · published May 26, 2020 · accessed Oct 21, 2026
  6. facebookresearch/detectron2 README (citation entry and license)Meta (GitHub repository) · Repository · accessed Oct 21, 2026
  7. Convolutional Neural Networks for joint object detection and pose estimation: A comparative study (arXiv 1412.7190)arXiv · Paper · published Dec 22, 2014 · accessed Oct 21, 2026
  8. Crafting a multi-task CNN for viewpoint estimation (arXiv 1609.03894)arXiv · Paper · published Sep 13, 2016 · accessed Oct 21, 2026
  9. Training data-efficient image transformers & distillation through attention (arXiv 2012.12877)arXiv · Paper · published Dec 23, 2020 · accessed Oct 21, 2026
  10. DINOv2: Learning Robust Visual Features without Supervision (arXiv 2304.07193)arXiv · Paper · published Apr 14, 2023 · accessed Oct 21, 2026
  11. Hybrid Transformers for Music Source Separation (arXiv 2211.08553)arXiv · Paper · published Nov 15, 2022 · accessed Oct 21, 2026
  12. CWM: An Open-Weights LLM for Research on Code Generation with World Models (arXiv 2510.02387)arXiv · Paper · published Sep 30, 2025 · accessed Oct 21, 2026
  13. PyTorch: An Imperative Style, High-Performance Deep Learning Library (arXiv 1912.01703)arXiv · Paper · published Dec 3, 2019 · accessed Oct 21, 2026
  14. facebookresearch/detr README (license)Meta (GitHub repository) · Repository · accessed Oct 21, 2026