Project AURIONAI Use-Cases Research & Innovation
Nanjing University
vision

VideoMAE

Self-supervised video pretraining with masked autoencoders, from Nanjing University.

Nanjing UniversityNanjing UniversityChinaTencentTencentChina
Diagram of the VideoMAE framework with tube masking and a high masking ratio, from the VideoMAE README.
Levelmodel
Classvision
Parameters94.21M
Contextnot in the notes
Released23 Mar 2022
Opennessopen weights restricted
Runs onmobile or NPU
Statusactive

Lineage

successor
VideoMAE to VideoMAE V2: successor, high confidenceVideoMAEVideoMAENanjing University, Mar 2022VideoMAE V2VideoMAE V2Mar 2023
VideoMAE V2 is a child (successor)high confidencenotes
Where this comes from
Parent and child relations from model cards, Hugging Face base_model fields and the genealogy tables of the research notes. The tree shows two generations each way, 2 models. Confidence and evidence are the notes' own.

R&D Topics

Video action recognition and temporal understandingproduced bySelf-supervised visual representation learningproduced by

Identity

Developer
Nanjing University, Tencent
Family
not in the notes
Released
23 Mar 2022 sourcesannouncement page
Weights published
3 Aug 2022
Record checked
12 Oct 2026
Where this comes from
Developer and dates from the research notes. Release date is the vendor's public announcement; weights date is when open weights were published.

Architecture and Size

Architecture
dense transformer
Total parameters
94.21M sources
Active parameters
–
Checkpoint parameters
–
Note
VideoMAE small, base, large; VideoMAE V2 (arXiv 2303.16727) up to ViT-g.
Context length
not in the notes
Input
video
Output
not in the notes
Where this comes from
Parameter counts as the developer states them. Where the checkpoint holds a different total, both are shown (register A16).

Openness and Access

Openness
open weights restricted
Licence
Creative Commons Attribution-NonCommercial 4.0 sources
Access routes
open weights download
Where this comes from
The licence file governs over registry tags and aggregator labels.

Deployment

Runs on
mobile or NPU
Footprint
Edge or toolkit builds listed in: Qualcomm AI Hub.
Runtimes
not in the notes
Where this comes from
Deployment tiers and runtimes as named in the notes. Footprint is the notes' figure, not a measurement by AURION.

Indonesia

Indonesian support
not stated
Regional languages
not in the notes
Jakarta hosting
unknown
Languages
not in the notes
Where this comes from
Indonesia fields are carried on every model; unknown is an allowed value.

Benchmark Results

The notes record no benchmark result for this model.

Industry Use Cases

No industry use case links this model.

Sources

  1. arxiv.org/abs/2203.12602 https://arxiv.org/abs/2203.12602
  2. huggingface.co/MCG-NJU/videomae-base https://huggingface.co/MCG-NJU/videomae-base
  3. aihub.qualcomm.com/models/video_mae https://aihub.qualcomm.com/models/video_mae
  4. github.com/MCG-NJU/VideoMAE https://github.com/MCG-NJU/VideoMAE
  5. huggingface.co/api/models?pipeline_tag=video-classification&... https://huggingface.co/api/models?pipeline_tag=video-classification&sort=downloads&limit=50
  6. github.com/open-mmlab/mmaction2/tree/main/configs/recogn... https://github.com/open-mmlab/mmaction2/tree/main/configs/recognition/videomae
  7. github.com/open-mmlab/mmaction2/tree/main/configs/recogn... https://github.com/open-mmlab/mmaction2/tree/main/configs/recognition/videomaev2
  8. arxiv.org/html/2203.12602 https://arxiv.org/html/2203.12602

To cite this page: Project AURION, AI Use-Cases Research & Innovation, "VideoMAE", https://aurion.semai.id/models/videomae/, records of October 2026.