1 code implementation • 11 Aug 2022 • Zejiang Hou, Fei Sun, Yen-Kuang Chen, Yuan Xie, Sun-Yuan Kung
When the masked autoencoder is pretrained and finetuned on ImageNet-1K dataset with an input resolution of 224x224, MILAN achieves a top-1 accuracy of 85. 4% on ViT-Base, surpassing previous state-of-the-arts by 1%.
1 code implementation • CVPR 2022 • Zejiang Hou, Minghai Qin, Fei Sun, Xiaolong Ma, Kun Yuan, Yi Xu, Yen-Kuang Chen, Rong Jin, Yuan Xie, Sun-Yuan Kung
However, conventional pruning methods have limitations in that: they are restricted to pruning process only, and they require a fully pre-trained large model.
1 code implementation • 31 Dec 2021 • Zejiang Hou, Sun-Yuan Kung
In contrast, we advocate a multi-dimensional ViT compression paradigm, and propose to harness the redundancy reduction from attention head, neuron and sequence dimensions jointly.
no code implementations • 7 Sep 2021 • Zejiang Hou, Sun-Yuan Kung
We study the few-shot learning (FSL) problem, where a model learns to recognize new objects with extremely few labeled training data per category.
no code implementations • 3 Jan 2021 • Tanvir Mahmud, A. Q. M. Sazzad Sayyed, Shaikh Anowarul Fattah, Sun-Yuan Kung
In this paper, we have proposed a novel multi-stage training approach that increases diversity in this feature extraction process to make accurate recognition of actions by combining varieties of features extracted from diverse perspectives.
no code implementations • 2 Dec 2020 • Tanvir Mahmud, Md Awsafur Rahman, Shaikh Anowarul Fattah, Sun-Yuan Kung
Moreover, a multi-scale fusion module is introduced with a pyramid fusion scheme to reduce the semantic gaps between subsequent encoder/decoder modules while facilitating the parallel optimization for efficient gradient propagation.
no code implementations • 30 Jul 2020 • Mert Al, Semih Yagli, Sun-Yuan Kung
The rapid rise of IoT and Big Data has facilitated copious data driven applications to enhance our quality of life.
no code implementations • 28 May 2020 • Zejiang Hou, Sun-Yuan Kung
Network pruning has become the de facto tool to accelerate deep neural networks for mobile and edge applications.
no code implementations • 1 Mar 2020 • Yuan Zhou, Dandan Li, Shuwei Huo, Sun-Yuan Kung
At present, the most effective and widely-used activation function is ReLU.
no code implementations • 24 Nov 2019 • Xukai Xie, Yuan Zhou, Sun-Yuan Kung
All the existing methods determine the importance of each operation directly by architecture weights.
no code implementations • 4 Nov 2019 • Yuan Zhou, Hongru Li, Sun-Yuan Kung
In the present study, we developed a novel method, referred to as Gemini Network, for effective modeling of temporal structures and achieving high-performance temporal action localization.
no code implementations • 4 Nov 2019 • Yuan Zhou, Xiaoting Du, Yeda Zhang, Sun-Yuan Kung
To this end, we propose the cross-scale residual network to exploit scale-related features and the inter-task correlations among the three tasks.
no code implementations • 1 Nov 2019 • Dandan Li, Yuan Zhou, Shuwei Huo, Sun-Yuan Kung
Convolutional neural networks (CNNs) are inherently suffering from massively redundant computation (FLOPs) due to the dense connection pattern between feature maps and convolution kernels.
no code implementations • 23 Sep 2019 • Mert Al, Zejiang Hou, Sun-Yuan Kung
Kernel approximation methods create explicit, low-dimensional kernel feature maps to deal with the high computational and memory complexity of standard techniques.
no code implementations • 9 Jun 2019 • Xukai Xie, Yuan Zhou, Sun-Yuan Kung
Using this operation, feature maps of different group cannot communicate, which restricts their representation capability.
no code implementations • 10 May 2018 • Mert Al, Thee Chanyaswad, Sun-Yuan Kung
They approximate explicit, low-dimensional feature mappings for kernel functions from the pairwise comparisons with the training data.
no code implementations • 8 Aug 2017 • Shibiao Wan, Man-Wai Mak, Sun-Yuan Kung
In the post-genomic era, large-scale personal DNA sequences are produced and collected for genetic medical diagnoses and new drug discovery, which, however, simultaneously poses serious challenges to the protection of personal genomic privacy.
no code implementations • 26 Feb 2017 • Mert Al, Shibiao Wan, Sun-Yuan Kung
With a rapidly increasing number of devices connected to the internet, big data has been applied to various domains of human life.
no code implementations • 29 Jan 2015 • Qi Guo, Bo-Wei Chen, Feng Jiang, Xiangyang Ji, Sun-Yuan Kung
Firstly, we divide the feature space into several subspaces using the decomposition method proposed in this paper.