Power Efficient Video Super-Resolution on Mobile NPUs with Deep Learning, Mobile AI & AIM 2022 Challenge: Report
Abstract
Video super-resolution is one of the most popular tasks on mobile devices, being widely used for an automatic improvement of low-bitrate and low-resolution video streams. While numerous solutions have been proposed for this problem, they are usually quite computationally demanding, demonstrating low FPS rates and power efficiency on mobile devices. In this Mobile AI challenge, we address this problem and propose the participants to design an end-to-end real-time video super-resolution solution for mobile NPUs optimized for low energy consumption. The participants were provided with the REDS training dataset containing video sequences for a 4X video upscaling task. The runtime and power efficiency of all models was evaluated on the powerful MediaTek Dimensity 9000 platform with a dedicated AI processing unit capable of accelerating floating-point and quantized neural networks. All proposed solutions are fully compatible with the above NPU, demonstrating an up to 500 FPS rate and 0.2 [Watt/30 FPS] power consumption. A detailed description of all models developed in the challenge is provided in this paper.
Cite
Text
Ignatov et al. "Power Efficient Video Super-Resolution on Mobile NPUs with Deep Learning, Mobile AI & AIM 2022 Challenge: Report." European Conference on Computer Vision Workshops, 2022. doi:10.1007/978-3-031-25066-8_6Markdown
[Ignatov et al. "Power Efficient Video Super-Resolution on Mobile NPUs with Deep Learning, Mobile AI & AIM 2022 Challenge: Report." European Conference on Computer Vision Workshops, 2022.](https://mlanthology.org/eccvw/2022/ignatov2022eccvw-power/) doi:10.1007/978-3-031-25066-8_6BibTeX
@inproceedings{ignatov2022eccvw-power,
title = {{Power Efficient Video Super-Resolution on Mobile NPUs with Deep Learning, Mobile AI & AIM 2022 Challenge: Report}},
author = {Ignatov, Andrey and Timofte, Radu and Chiang, Cheng-Ming and Kuo, Hsien-Kai and Xu, Yu-Syuan and Lee, Man-Yu and Lu, Allen and Cheng, Chia-Ming and Chen, Chih-Cheng and Yong, Jia-Ying and Shuai, Hong-Han and Cheng, Wen-Huang and Jia, Zhuang and Xu, Tianyu and Zhang, Yijian and Bao, Long and Sun, Heng and Zhang, Diankai and Gao, Si and Liu, Shaoli and Wu, Biao and Zhang, Xiaofeng and Zheng, Chengjian and Lu, Kaidi and Wang, Ning and Sun, Xiao and Wu, Haodong and Liu, Xuncheng and Zhang, Weizhan and Yan, Caixia and Du, Haipeng and Zheng, Qinghua and Wang, Qi and Chen, Wangdu and Duan, Ran and Sun, Mengdi and Zhu, Dan and Chen, Guannan and Cho, Hojin and Kim, Steve and Yue, Shijie and Li, Chenghua and Zhuge, Zhengyang and Chen, Wei and Wang, Wenxu and Zhou, Yufeng and Cai, Xiaochen and Cai, Hengxing and Xu, Kele and Liu, Li and Cheng, Zehua and Lian, Wenyi and Lian, Wenjing},
booktitle = {European Conference on Computer Vision Workshops},
year = {2022},
pages = {130-152},
doi = {10.1007/978-3-031-25066-8_6},
url = {https://mlanthology.org/eccvw/2022/ignatov2022eccvw-power/}
}