Researcher

Mohamed Omran

Mohamed Omran is a computer vision researcher who coauthored Cityscapes, a street-scene dataset that became a go-to benchmark for segmentation and detection. They studied at Ulm University and Saarland University, where they earned an M.Sc. in Visual Computing in 2014, and then began a PhD at the Max Planck Institute for Informatics. Work from those years includes "Ten Years of Pedestrian Detection, What Have We Learned?" in 2014, the Cityscapes paper in 2016, and Neural Body Fitting, which won Best Student Paper at 3DV 2018. The 2021 thesis "From Pixels to People" drew together their pedestrian detection, Cityscapes and 3D human shape and pose work. Since 2022, they are a researcher at Qualcomm, and in 2025 they coauthored the Qualcomm AI Research entry that placed second in the RealADSim-NVS challenge.

Now at
Qualcomm
Badges

Timeline

  1. 2010–2014

    Studies at Ulm and Saarland

    They work as a research assistant at Daimler Research and Development in Ulm in 2010/11 and earn a B.Sc. in Media Informatics at Ulm University in 2011. In 2014, they earn an M.Sc. in Visual Computing at Saarland University.

  2. 2014

    Ten Years of Pedestrian Detection

    In 2014, they start a PhD in computer science at the Max Planck Institute for Informatics and, with Rodrigo Benenson, Jan Hosang and Bernt Schiele, coauthor "Ten Years of Pedestrian Detection, What Have We Learned?". The paper reviews the 40+ detectors on the Caltech benchmark and combines published strategies into a decision forest detector with the best known performance on Caltech-USA.

  3. 2016

    The Cityscapes dataset

    In April 2016, they coauthor the Cityscapes paper with Marius Cordts, Bernt Schiele and others. Cityscapes offers stereo video from 50 cities, with 5000 finely annotated and 20000 coarsely annotated images, and goes on to become a go-to benchmark for segmentation and detection.

  4. 2018

    Neural Body Fitting at 3DV

    In 2018, they publish Neural Body Fitting with Christoph Lassner, Gerard Pons-Moll, Peter Gehler and Bernt Schiele. The method for 3D human pose and shape estimation integrates a statistical body model within a CNN, and the paper wins the Best Student Paper award at 3DV 2018.

  5. 2021

    PhD thesis From Pixels to People

    They complete the PhD thesis "From Pixels to People" at Universität des Saarlandes in Saarbrücken in 2021. It covers pedestrian detection, the Cityscapes dataset, boundary detection and a method for recovering 3D human shape and pose.

  6. Since 2022

    Researcher at Qualcomm

    They join Qualcomm as a researcher in 2022.

  7. October 2024

    Generative object placement

    In October 2024, they coauthor "Generative Location Modeling for Spatially Aware Object Insertion" with Jooyeol Yun, Davide Abati, Jaegul Choo, Amirhossein Habibian and Auke Wiggers. The paper, from Qualcomm AI Research, trains an autoregressive model that generates bounding box coordinates for realistic object locations.

  8. October 2025

    Second place at RealADSim-NVS

    They coauthor "Hybrid Gaussian Splatting for Novel Urban View Synthesis" with Farhad Zanjani, Davide Abati, Jens Petersen and Amirhossein Habibian. The paper describes the Qualcomm AI Research solution to the RealADSim-NVS challenge at ICCV 2025, which reaches an aggregated score of 0.432 and second place on the public leaderboard.

Career

  1. Since 2022QualcommResearcher
  2. 2010–2011DaimlerResearch assistant

Education

  1. 2021Saarland UniversityPhD
  2. 2014Saarland UniversityM.Sc.
  3. From 2014Max Planck Institute for InformaticsPhD
  4. 2011Universität UlmB.Sc.

Sources and credits

Photo: We have no published photo of Mohamed Omran yet. Send us a free-licensed photo.

  1. Mohamed Omran - Max Planck Institute for Informatics mpi-inf.mpg.de. Checked 21 October 2026.
  2. [1411.4304] Ten Years of Pedestrian Detection, What Have We Learned? arxiv.org. Checked 21 October 2026.
  3. [1604.01685] The Cityscapes Dataset for Semantic Urban Scene Understanding arxiv.org. Checked 21 October 2026.
  4. International Conference on 3D Vision 2018 3dv18.uniud.it. Checked 21 October 2026.
  5. [1808.05942] Neural Body Fitting: Unifying Deep Learning and Model-Based Human Pose and Shape Estimation arxiv.org. Checked 21 October 2026.
  6. Mohamed Omran (Qualcomm): computer vision, machine learning lacuna.tiptreesystems.com. Checked 21 October 2026.
  7. Generative Location Modeling for Spatially Aware Object Insertion arxiv.org. Checked 21 October 2026.
  8. [2510.12308] Hybrid Gaussian Splatting for Novel Urban View Synthesis arxiv.org. Checked 21 October 2026.