PeopleofSuperintelligence

Researcher

Junting Pan

Also known as 潘俊廷

Junting Pan is a computer vision researcher whose work runs from visual saliency and video understanding to multimodal foundation models. In 2015, his network JuntingNet, built with Xavier Giró-i-Nieto at the Universitat Politècnica de Catalunya, won the saliency prediction task of the LSUN challenge, and in 2017 he was first author of SalGAN. He did his PhD in the Multimedia Lab at the Chinese University of Hong Kong under Xiaogang Wang and Hongsheng Li, interned at Meta FAIR in 2023, and was among the 18 authors of SAM 2 in 2024. SAM 2 received an Honorable Mention among the ICLR 2025 Outstanding Paper awards. He is a research scientist on the Foundation Models Team at Apple AIML.

Now at
Apple

Timeline

  1. 2015

    JuntingNet wins LSUN 2015

    In 2015, he and Xavier Giró-i-Nieto of the Image Processing Group at the Universitat Politècnica de Catalunya (UPC) win the saliency prediction task of the Large-scale Scene Understanding (LSUN) challenge at CVPR 2015. Their network, JuntingNet, is trained end to end on large saliency datasets and shows superior performance on all considered metrics.

  2. 2017

    SalGAN and adversarial saliency

    In 2017, he is first author of SalGAN, a deep convolutional network for visual saliency prediction trained with adversarial examples. A generator learns to produce saliency maps, and a discriminator is trained to tell them from the ground-truth maps; the coauthors include Cristian Canton Ferrer, Kevin McGuinness and Xavier Giró-i-Nieto.

  3. 2020–2021

    Actor-Context-Actor Relation Network

    As a PhD student in the Multimedia Lab (MMLab) at the Chinese University of Hong Kong, supervised by Xiaogang Wang and Hongsheng Li, he proposes the Actor-Context-Actor Relation Network for spatio-temporal action localization. The method ranks first in the AVA-Kinetics action localization task of the ActivityNet Challenge 2020 and is published at CVPR 2021.

  4. 2022

    ST-Adapter and EdgeViTs

    In 2022, he is first author of the ST-Adapter, a Spatio-Temporal Adapter that lets a pre-trained image model reason about dynamic video content through parameter-efficient fine-tuning; the paper is accepted to NeurIPS 2022. He is also first author of EdgeViTs, a family of light-weight vision transformers that compete with light-weight CNNs on mobile devices, published at ECCV 2022.

  5. 2023

    Meta FAIR and JourneyDB

    In May 2023, he starts an internship as a research scientist intern at Meta AI (FAIR). He also coauthors JourneyDB, a benchmark of 4 million generated images paired with their text prompts for generative image understanding, accepted to the NeurIPS 2023 Datasets and Benchmarks track.

  6. 2024

    SAM 2 and Math-Vision

    In 2024, he is among the 18 authors of SAM 2, a foundation model for promptable visual segmentation in images and videos that is 6x faster than the Segment Anything Model on images and needs 3x fewer interactions in video. He also coauthors Math-Vision, a collection of 3,040 mathematical problems with visual contexts from real math competitions, accepted to the NeurIPS 2024 Datasets and Benchmarks track.

  7. 2025

    An ICLR 2025 honorable mention

    In 2025, the ICLR Outstanding Paper committee names SAM 2: Segment Anything in Images and Videos one of three Honorable Mentions. He is one of its 18 authors.

  8. Since 2025

    Research scientist at Apple

    He is a research scientist on the Foundation Models Team at Apple AIML. His research interests lie in multimodal foundation models, particularly vision-language alignment, video understanding and generative tasks. In July 2025, he appears among the authors of the Apple Intelligence Foundation Language Models Tech Report.

Career

  1. Since 2025AppleResearch Scientist
  2. From 2023MetaResearch Scientist Intern

Education

  1. Chinese University of Hong KongPhD

From the biography.

  1. Google alumni have founded or led 133 other organizations, more than any other AI labIncludes Junting Pan. We estimate that alumni of 17 major AI labs went on to found or lead 350 other organizations, including 226 they founded, and we trace 472 career moves directly between 23 labs since 2010.
  2. Of 943 people shaping AI who worked at Google, 446 work there nowIncludes Apple. Our estimates of where the 943 people shaping AI who worked at Google and Google DeepMind work now: 446 at Google and 452 across 278 other organizations.
  3. Among employers with 30+ full-time cases, OpenAI leads primary technical role H-1B pay at $310,000, and Amazon, Google and Microsoft file 57% of all casesIncludes Apple. Across 440,669 certified US H-1B filings from 44 employers (FY2020–FY2026), OpenAI leads primary technical role base pay at $310,000 and xAI leads overall base pay at $290,000.
  4. The top 10 organizations employed 22% of people shaping AI in 2025: 51% of company researchers and 21% of university researchersIncludes Apple. We estimate the share of people shaping AI who worked at the top 10 organizations each year from 1955 to 2025, comparing concentration across companies, universities, and institutes and labs.

Colleagues

People with dated positions at the same organization in the same years.

Sources and credits

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

  1. Success at CVPR workshops: Junting Pan and Xavier Giro awarded in the LSUN challenge. ChaLearn prize also collected | Image Processing Group imatge.upc.edu. Checked 21 October 2026.
  2. End-to-end Convolutional Network for Saliency Prediction (arXiv:1507.01422) arxiv.org. Checked 21 October 2026.
  3. End-to-end Convolutional Networks for Saliency Prediction (software) | Image Processing Group, UPC imatge.upc.edu. Checked 21 October 2026.
  4. SalGAN: Visual Saliency Prediction with Generative Adversarial Networks (arXiv:1701.01081) arxiv.org. Checked 21 October 2026.
  5. Junting Pan junting.github.io. Checked 21 October 2026.
  6. ST-Adapter: Parameter-Efficient Image-to-Video Transfer Learning (arXiv:2206.13559) arxiv.org. Checked 21 October 2026.
  7. JourneyDB: A Benchmark for Generative Image Understanding (arXiv:2307.00716) arxiv.org. Checked 21 October 2026.
  8. SAM 2: Segment Anything in Images and Videos (arXiv:2408.00714) arxiv.org. Checked 21 October 2026.
  9. Announcing the Outstanding Paper Awards at ICLR 2025 | ICLR Blog blog.iclr.cc. Checked 21 October 2026.
  10. Apple Intelligence Foundation Language Models: Tech Report 2025 (arXiv:2507.13575) arxiv.org. Checked 21 October 2026.