1 code implementation • 14 Mar 2024 • Tao Huang, Xiaohuan Pei, Shan You, Fei Wang, Chen Qian, Chang Xu
This paper posits that the key to enhancing Vision Mamba (ViM) lies in optimizing scan directions for sequence modeling.
no code implementations • 11 Mar 2024 • Tao Huang, Jiaqi Liu, Shan You, Chang Xu
Recently, the growing capabilities of deep generative models have underscored their potential in enhancing image classification accuracy.
1 code implementation • NeurIPS 2023 • Yichao Cao, Qingfei Tang, Xiu Su, Chen Song, Shan You, Xiaobo Lu, Chang Xu
We conduct a deep analysis of the three hierarchical features inherent in visual HOI detectors and propose a method for high-level relation extraction aimed at VL foundation models, which we call HO prompt-based learning.
1 code implementation • 21 Aug 2023 • Mingkai Zheng, Shan You, Lang Huang, Xiu Su, Fei Wang, Chen Qian, Xiaogang Wang, Chang Xu
Moreover, to further boost the performance, we propose ``distributional consistency" as a more informative regularization to enable similar instances to have a similar probability distribution.
2 code implementations • ICCV 2023 • Mingkai Zheng, Shan You, Lang Huang, Chen Luo, Fei Wang, Chen Qian, Chang Xu
Semi-Supervised image classification is one of the most fundamental problem in computer vision, which significantly reduces the need for human labor.
no code implementations • ICCV 2023 • Yichao Cao, Qingfei Tang, Feng Yang, Xiu Su, Shan You, Xiaobo Lu, Chang Xu
Human-Object Interaction (HOI) detection is a challenging computer vision task that requires visual models to address the complex interactive relationship between humans and objects and predict HOI triplets.
1 code implementation • NeurIPS 2023 • Tao Huang, Yuan Zhang, Mingkai Zheng, Shan You, Fei Wang, Chen Qian, Chang Xu
To address this, we propose to denoise student features using a diffusion model trained by teacher features.
1 code implementation • 21 Apr 2023 • Mingkai Zheng, Xiu Su, Shan You, Fei Wang, Chen Qian, Chang Xu, Samuel Albanie
We investigate the potential of GPT-4~\cite{gpt4} to perform Neural Architecture Search (NAS) -- the task of designing effective neural architectures.
1 code implementation • 26 Oct 2022 • Haoyu Xie, Changqi Wang, Mingkai Zheng, Minjing Dong, Shan You, Chong Fu, Chang Xu
In prevalent pixel-wise contrastive learning solutions, the model maps pixels to deterministic representations and regularizes them in the latent space.
1 code implementation • 15 Jul 2022 • Jiyang Xie, Xiu Su, Shan You, Zhanyu Ma, Fei Wang, Chen Qian
Recently, community has paid increasing attention on model scaling and contributed to developing a model family with a wide spectrum of scales.
1 code implementation • 12 Jul 2022 • Tao Huang, Lang Huang, Shan You, Fei Wang, Chen Qian, Chang Xu
Vision transformers (ViTs) are usually considered to be less light-weight than convolutional neural networks (CNNs) due to the lack of inductive bias.
1 code implementation • 29 May 2022 • Tao Huang, Yuan Zhang, Shan You, Fei Wang, Chen Qian, Jian Cao, Chang Xu
To obtain a group of masks, the receptive tokens are learned via the regular task loss but with teacher fixed, and we also leverage a Dice loss to enrich the diversity of learned masks.
1 code implementation • 26 May 2022 • Lang Huang, Shan You, Mingkai Zheng, Fei Wang, Chen Qian, Toshihiko Yamasaki
We present an efficient approach for Masked Image Modeling (MIM) with hierarchical Vision Transformers (ViTs), allowing the hierarchical ViTs to discard masked patches and operate only on the visible ones.
2 code implementations • 21 May 2022 • Tao Huang, Shan You, Fei Wang, Chen Qian, Chang Xu
In this paper, we show that simply preserving the relations between the predictions of teacher and student would suffice, and propose a correlation-based loss to capture the intrinsic inter-class relations from the teacher explicitly.
Ranked #2 on Knowledge Distillation on ImageNet (using extra training data)
1 code implementation • CVPR 2022 • Lang Huang, Shan You, Mingkai Zheng, Fei Wang, Chen Qian, Toshihiko Yamasaki
In this paper, we present a new approach, Learning Where to Learn (LEWEL), to adaptively aggregate spatial information of features, so that the projected embeddings could be exactly aligned and thus guide the feature learning better.
1 code implementation • 25 Mar 2022 • Xiu Su, Shan You, Jiyang Xie, Fei Wang, Chen Qian, ChangShui Zhang, Chang Xu
In BCNet, each channel is fairly trained and responsible for the same amount of network widths, thus each network width can be evaluated more accurately.
2 code implementations • CVPR 2022 • Tao Huang, Shan You, Bohan Zhang, Yuxuan Du, Fei Wang, Chen Qian, Chang Xu
Structural re-parameterization (Rep) methods achieve noticeable improvements on simple VGG-style networks.
no code implementations • 16 Mar 2022 • Mingkai Zheng, Shan You, Fei Wang, Chen Qian, ChangShui Zhang, Xiaogang Wang, Chang Xu
Self-supervised Learning (SSL) including the mainstream contrastive learning has achieved great success in learning visual representations without data annotations.
Ranked #60 on Self-Supervised Image Classification on ImageNet
1 code implementation • CVPR 2022 • Mingkai Zheng, Shan You, Lang Huang, Fei Wang, Chen Qian, Chang Xu
Learning with few labeled data has been a longstanding problem in the computer vision and machine learning research community.
1 code implementation • ICLR 2022 • Tao Huang, Zekang Li, Hua Lu, Yong Shan, Shusheng Yang, Yang Feng, Fei Wang, Shan You, Chang Xu
Evaluation metrics in machine learning are often hardly taken as loss functions, as they could be non-differentiable and non-decomposable, e. g., average precision and F1 score.
no code implementations • CVPR 2022 • Tao Huang, Shan You, Fei Wang, Chen Qian, ChangShui Zhang, Xiaogang Wang, Chang Xu
In this paper, we leverage an explicit path filter to capture the characteristics of paths and directly filter those weak ones, so that the search can be thus implemented on the shrunk space more greedily and efficiently.
1 code implementation • ICCV 2021 • Mingkai Zheng, Fei Wang, Shan You, Chen Qian, ChangShui Zhang, Xiaogang Wang, Chang Xu
Specifically, our proposed framework is based on two projection heads, one of which will perform the regular instance discrimination task.
2 code implementations • NeurIPS 2021 • Mingkai Zheng, Shan You, Fei Wang, Chen Qian, ChangShui Zhang, Xiaogang Wang, Chang Xu
Self-supervised Learning (SSL) including the mainstream contrastive learning has achieved great success in learning visual representations without data annotations.
Ranked #78 on Self-Supervised Image Classification on ImageNet
1 code implementation • 25 Jun 2021 • Xiu Su, Shan You, Jiyang Xie, Mingkai Zheng, Fei Wang, Chen Qian, ChangShui Zhang, Xiaogang Wang, Chang Xu
Vision transformers (ViTs) inherited the success of NLP but their structures have not been sufficiently investigated and optimized for visual tasks.
no code implementations • 11 Jun 2021 • Xiu Su, Shan You, Mingkai Zheng, Fei Wang, Chen Qian, ChangShui Zhang, Chang Xu
The operation weight for each path is represented as a convex combination of items in a dictionary with a simplex code.
no code implementations • CVPR 2021 • Xiu Su, Shan You, Fei Wang, Chen Qian, ChangShui Zhang, Chang Xu
In BCNet, each channel is fairly trained and responsible for the same amount of network widths, thus each network width can be evaluated more accurately.
1 code implementation • CVPR 2021 • Xiu Su, Tao Huang, Yanxi Li, Shan You, Fei Wang, Chen Qian, ChangShui Zhang, Chang Xu
One-shot neural architecture search (NAS) methods significantly reduce the search cost by considering the whole search space as one network, which only needs to be trained once.
no code implementations • ICLR 2021 • Xiu Su, Shan You, Tao Huang, Fei Wang, Chen Qian, ChangShui Zhang, Chang Xu
In this paper, to better evaluate each width, we propose a locally free weight sharing strategy (CafeNet) accordingly.
no code implementations • CVPR 2021 • Yibo Yang, Shan You, Hongyang Li, Fei Wang, Chen Qian, Zhouchen Lin
Our method enables differentiable sparsification, and keeps the derived architecture equivalent to that of Engine-cell, which further improves the consistency between search and evaluation.
no code implementations • ICCV 2021 • Yuru Song, Zan Lou, Shan You, Erkun Yang, Fei Wang, Chen Qian, ChangShui Zhang, Xiaogang Wang
Concretely, we introduce a privileged parameter so that the optimization direction does not necessarily follow the gradient from the privileged tasks, but concentrates more on the target tasks.
no code implementations • 1 Jan 2021 • Zhuozhuo Tu, Shan You, Tao Huang, DaCheng Tao
Wasserstein distributionally robust optimization (DRO) has recently received significant attention in machine learning due to its connection to generalization, robustness and regularization.
no code implementations • 1 Jan 2021 • Yibo Yang, Shan You, Hongyang Li, Fei Wang, Chen Qian, Zhouchen Lin
The Engine-cell is differentiable for architecture search, while the Transit-cell only transits the current sub-graph by architecture derivation.
no code implementations • 1 Jan 2021 • Tao Huang, Shan You, Yibo Yang, Zhuozhuo Tu, Fei Wang, Chen Qian, ChangShui Zhang
Differentiable neural architecture search (NAS) has gained much success in discovering more flexible and diverse cell types.
1 code implementation • NeurIPS 2020 • Shangchen Du, Shan You, Xiaojie Li, Jianlong Wu, Fei Wang, Chen Qian, ChangShui Zhang
In this paper, we examine the diversity of teacher models in the gradient space and regard the ensemble knowledge distillation as a multi-objective optimization problem so that we can determine a better optimization direction for the training of student network.
no code implementations • 18 Nov 2020 • Tao Huang, Shan You, Yibo Yang, Zhuozhuo Tu, Fei Wang, Chen Qian, ChangShui Zhang
However, even for this consistent search, the searched cells often suffer from poor performance, especially for the supernet with fewer layers, as current DARTS methods are prone to wide and shallow cells, and this topology collapse induces sub-optimal searched cells.
no code implementations • 28 Oct 2020 • Xiu Su, Shan You, Tao Huang, Hongyan Xu, Fei Wang, Chen Qian, ChangShui Zhang, Chang Xu
To deploy a well-trained CNN model on low-end computation edge devices, it is usually supposed to compress or prune the model under certain computation budget (e. g., FLOPs).
1 code implementation • 20 Oct 2020 • Yuxuan Du, Tao Huang, Shan You, Min-Hsiu Hsieh, DaCheng Tao
Variational quantum algorithms (VQAs) are expected to be a path to quantum advantages on noisy intermediate-scale quantum devices.
1 code implementation • NeurIPS 2020 • Yibo Yang, Hongyang Li, Shan You, Fei Wang, Chen Qian, Zhouchen Lin
By doing so, our network for search at each update satisfies the sparsity constraint and is efficient to train.
no code implementations • 23 Jul 2020 • Yuxuan Du, Min-Hsiu Hsieh, Tongliang Liu, Shan You, DaCheng Tao
The eligibility of various advanced quantum algorithms will be questioned if they can not guarantee privacy.
no code implementations • CVPR 2020 • Shan You, Tao Huang, Mingmin Yang, Fei Wang, Chen Qian, Chang-Shui Zhang
The training efficiency is thus boosted since the training space has been greedily shrunk from all paths to those potentially-good ones.
Ranked #72 on Neural Architecture Search on ImageNet
no code implementations • 13 Jul 2019 • Yehui Tang, Shan You, Chang Xu, Boxin Shi, Chao Xu
Specifically, we exploit the unlabeled data to mimic the classification characteristics of giant networks, so that the original capacity can be preserved nicely.
no code implementations • 25 Jan 2017 • Shan You, Chang Xu, Yunhe Wang, Chao Xu, DaCheng Tao
This paper presents privileged multi-label learning (PrML) to explore and exploit the relationship between labels in multi-label learning problems.
no code implementations • NeurIPS 2016 • Yunhe Wang, Chang Xu, Shan You, DaCheng Tao, Chao Xu
Deep convolutional neural networks (CNNs) are successfully used in a number of applications.
no code implementations • 19 Apr 2016 • Yunhe Wang, Chang Xu, Shan You, DaCheng Tao, Chao Xu
Here we study the extreme visual recovery problem, in which over 90\% of pixel values in a given image are missing.
no code implementations • 19 Apr 2016 • Shan You, Chang Xu, Yunhe Wang, Chao Xu, DaCheng Tao
The core of SLL is to explore and exploit the relationships between new labels and past labels and then inherit the relationship into hypotheses of labels to boost the performance of new classifiers.