VideoMAEv2-Huge 模型以自监督方式在 UnlabeldHybrid-1M 数据集上预训练了 1200 个 epoch。该模型由 Wang 等人在论文《[CVPR23]VideoMAE V2: Scaling Video Masked Autoencoders with Dual Masking》（https://arxiv.org/abs/2203.12602）中提出，并首次在 GitHub 上发布。
您可以使用原始模型进行视频特征提取。
以下是使用此模型提取视频特征的方法：
from transformers import VideoMAEImageProcessor, AutoModel, AutoConfig
import numpy as np
import torch
config = AutoConfig.from_pretrained("OpenGVLab/VideoMAEv2-Huge", trust_remote_code=True)
processor = VideoMAEImageProcessor.from_pretrained("OpenGVLab/VideoMAEv2-Huge")
model = AutoModel.from_pretrained('OpenGVLab/VideoMAEv2-Huge', config=config, trust_remote_code=True)
video = list(np.random.rand(16, 3, 224, 224))
# B, T, C, H, W -> B, C, T, H, W
inputs = processor(video, return_tensors="pt")
inputs['pixel_values'] = inputs['pixel_values'].permute(0, 2, 1, 3, 4)
with torch.no_grad():
outputs = model(**inputs)@InProceedings{wang2023videomaev2,
author = {Wang, Limin and Huang, Bingkun and Zhao, Zhiyu and Tong, Zhan and He, Yinan and Wang, Yi and Wang, Yali and Qiao, Yu},
title = {VideoMAE V2: Scaling Video Masked Autoencoders With Dual Masking},
booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
month = {June},
year = {2023},
pages = {14549-14560}
}
@misc{videomaev2,
title={VideoMAE V2: Scaling Video Masked Autoencoders with Dual Masking},
author={Limin Wang and Bingkun Huang and Zhiyu Zhao and Zhan Tong and Yinan He and Yi Wang and Yali Wang and Yu Qiao},
year={2023},
eprint={2303.16727},
archivePrefix={arXiv},
primaryClass={cs.CV}
}