Search Results for author: Wanli Ouyang

Found 215 papers, 96 papers with code

Boosting Residual Networks with Group Knowledge

no code implementations26 Aug 2023 Shengji Tang, Peng Ye, Baopu Li, Weihao Lin, Tao Chen, Tong He, Chong Yu, Wanli Ouyang

Specifically, we implicitly divide all subnets into hierarchical groups by subnet-in-subnet sampling, aggregate the knowledge of different subnets in each group during training, and exploit upper-level group knowledge to supervise lower-level subnet groups.

Knowledge Distillation

STEERER: Resolving Scale Variations for Counting and Localization via Selective Inheritance Learning

1 code implementation21 Aug 2023 Tao Han, Lei Bai, Lingbo Liu, Wanli Ouyang

Scale variation is a deep-rooted problem in object counting, which has not been effectively addressed by existing scale-aware algorithms.

feature selection Object Counting

Masked Motion Predictors are Strong 3D Action Representation Learners

1 code implementation14 Aug 2023 Yunyao Mao, Jiajun Deng, Wengang Zhou, Yao Fang, Wanli Ouyang, Houqiang Li

To be specific, the proposed MAMP takes as input the masked spatio-temporal skeleton sequence and predicts the corresponding temporal motion of the masked human joints.

3D Action Recognition motion prediction

Experts Weights Averaging: A New General Training Scheme for Vision Transformers

no code implementations11 Aug 2023 Yongqi Huang, Peng Ye, Xiaoshui Huang, Sheng Li, Tao Chen, Tong He, Wanli Ouyang

As Vision Transformers (ViTs) are gradually surpassing CNNs in various visual tasks, one may question: if a training scheme specifically for ViTs exists that can also achieve performance improvement without increasing inference cost?

MCTformer+: Multi-Class Token Transformer for Weakly Supervised Semantic Segmentation

1 code implementation6 Aug 2023 Lian Xu, Mohammed Bennamoun, Farid Boussaid, Hamid Laga, Wanli Ouyang, Dan Xu

Building upon the observation that the attended regions of the one-class token in the standard vision transformer can contribute to a class-agnostic localization map, we explore the potential of the transformer model to capture class-specific attention for class-discriminative object localization by learning multiple class tokens.

Object Localization Weakly supervised Semantic Segmentation +1

Theoretically Guaranteed Policy Improvement Distilled from Model-Based Planning

no code implementations24 Jul 2023 Chuming Li, Ruonan Jia, Jie Liu, Yinmin Zhang, Yazhe Niu, Yaodong Yang, Yu Liu, Wanli Ouyang

Model-based reinforcement learning (RL) has demonstrated remarkable successes on a range of continuous control tasks due to its high sample efficiency.

Continuous Control Model-based Reinforcement Learning +1

Meta-Transformer: A Unified Framework for Multimodal Learning

1 code implementation20 Jul 2023 Yiyuan Zhang, Kaixiong Gong, Kaipeng Zhang, Hongsheng Li, Yu Qiao, Wanli Ouyang, Xiangyu Yue

Multimodal learning aims to build models that can process and relate information from multiple modalities.

Time Series

What Can Simple Arithmetic Operations Do for Temporal Modeling?

1 code implementation18 Jul 2023 Wenhao Wu, Yuxin Song, Zhun Sun, Jingdong Wang, Chang Xu, Wanli Ouyang

We conduct comprehensive ablation studies on the instantiation of ATMs and demonstrate that this module provides powerful temporal modeling capability at a low computational cost.

Action Classification Action Recognition +1

MotionGPT: Finetuned LLMs are General-Purpose Motion Generators

no code implementations19 Jun 2023 Yaqi Zhang, Di Huang, Bin Liu, Shixiang Tang, Yan Lu, Lu Chen, Lei Bai, Qi Chu, Nenghai Yu, Wanli Ouyang

Generating realistic human motion from given action descriptions has experienced significant advancements because of the emerging requirement of digital humans.

UniG3D: A Unified 3D Object Generation Dataset

no code implementations19 Jun 2023 Qinghong Sun, Yangguang Li, Zexiang Liu, Xiaoshui Huang, Fenggang Liu, Xihui Liu, Wanli Ouyang, Jing Shao

However, the quality and diversity of existing 3D object generation methods are constrained by the inadequacies of existing 3D object datasets, including issues related to text quality, the incompleteness of multi-modal data representation encompassing 2D rendered images and 3D assets, as well as the size of the dataset.

Autonomous Driving

Adaptive Hierarchical SpatioTemporal Network for Traffic Forecasting

no code implementations15 Jun 2023 YiRong Chen, Ziyue Li, Wanli Ouyang, Michael Lepech

In this work, we propose an Adaptive Hierarchical SpatioTemporal Network (AHSTN) to promote traffic forecasting by exploiting the spatial hierarchy and modeling multi-scale spatial correlations.

Retrieve Anyone: A General-purpose Person Re-identification Task with Instructions

no code implementations13 Jun 2023 Weizhen He, Shixiang Tang, Yiheng Deng, Qihao Chen, Qingsong Xie, Yizhou Wang, Lei Bai, Feng Zhu, Rui Zhao, Wanli Ouyang, Donglian Qi, Yunfeng Yan

This paper strives to resolve this problem by proposing a new instruct-ReID task that requires the model to retrieve images according to the given image or language instructions. Our instruct-ReID is a more general ReID setting, where existing ReID tasks can be viewed as special cases by designing different instructions.

Person Re-Identification

LAMM: Language-Assisted Multi-Modal Instruction-Tuning Dataset, Framework, and Benchmark

1 code implementation11 Jun 2023 Zhenfei Yin, Jiong Wang, JianJian Cao, Zhelun Shi, Dingning Liu, Mukai Li, Lu Sheng, Lei Bai, Xiaoshui Huang, Zhiyong Wang, Jing Shao, Wanli Ouyang

2) We demonstrate the detailed methods of constructing instruction-tuning datasets and benchmarks for MLLMs, which will enable future research on MLLMs to scale up and extend to other domains, tasks, and modalities faster.

Bi-LRFusion: Bi-Directional LiDAR-Radar Fusion for 3D Dynamic Object Detection

1 code implementation CVPR 2023 Yingjie Wang, Jiajun Deng, Yao Li, Jinshui Hu, Cong Liu, Yu Zhang, Jianmin Ji, Wanli Ouyang, Yanyong Zhang

LiDAR and Radar are two complementary sensing approaches in that LiDAR specializes in capturing an object's 3D shape while Radar provides longer detection ranges as well as velocity hints.

object-detection Object Detection

PVT-SSD: Single-Stage 3D Object Detector with Point-Voxel Transformer

1 code implementation CVPR 2023 Honghui Yang, Wenxiao Wang, Minghao Chen, Binbin Lin, Tong He, Hua Chen, Xiaofei He, Wanli Ouyang

The key to associating the two different representations is our introduced input-dependent Query Initialization module, which could efficiently generate reference points and content queries.

Autonomous Driving Quantization

Stimulative Training++: Go Beyond The Performance Limits of Residual Networks

no code implementations4 May 2023 Peng Ye, Tong He, Shengji Tang, Baopu Li, Tao Chen, Lei Bai, Wanli Ouyang

In this work, we aim to re-investigate the training process of residual networks from a novel social psychology perspective of loafing, and further propose a new training scheme as well as three improved strategies for boosting residual networks beyond their performance limits.

Seeing is not always believing: Benchmarking Human and Model Perception of AI-Generated Images

1 code implementation25 Apr 2023 Zeyu Lu, Di Huang, Lei Bai, Jingjing Qu, Chengyue Wu, Xihui Liu, Wanli Ouyang

Along with this, we conduct the model capability of AI-Generated images detection evaluation MPBench and the top-performing model from MPBench achieves a 13% failure rate under the same setting used in the human evaluation.

Benchmarking Fake Image Detection

Automatically Predict Material Properties with Microscopic Image Example Polymer Compatibility

no code implementations22 Mar 2023 Zhilong Liang, Zhenzhi Tan, Ruixin Hong, Wanli Ouyang, Jinying Yuan, ChangShui Zhang

Computer image recognition with machine learning method can make up the defects of artificial judging, giving accurate and quantitative judgement.

Transfer Learning

HumanBench: Towards General Human-centric Perception with Projector Assisted Pretraining

1 code implementation CVPR 2023 Shixiang Tang, Cheng Chen, Qingsong Xie, Meilin Chen, Yizhou Wang, Yuanzheng Ci, Lei Bai, Feng Zhu, Haiyang Yang, Li Yi, Rui Zhao, Wanli Ouyang

Specifically, we propose a \textbf{HumanBench} based on existing datasets to comprehensively evaluate on the common ground the generalization abilities of different pretraining methods on 19 datasets from 6 diverse downstream tasks, including person ReID, pose estimation, human parsing, pedestrian attribute recognition, pedestrian detection, and crowd counting.

Autonomous Driving Crowd Counting +4

UniHCP: A Unified Model for Human-Centric Perceptions

1 code implementation CVPR 2023 Yuanzheng Ci, Yizhou Wang, Meilin Chen, Shixiang Tang, Lei Bai, Feng Zhu, Rui Zhao, Fengwei Yu, Donglian Qi, Wanli Ouyang

When adapted to a specific task, UniHCP achieves new SOTAs on a wide range of human-centric tasks, e. g., 69. 8 mIoU on CIHP for human parsing, 86. 18 mA on PA-100K for attribute prediction, 90. 3 mAP on Market1501 for ReID, and 85. 8 JI on CrowdHuman for pedestrian detection, performing better than specialized models tailored for each task.

2D Pose Estimation Human Parsing +7

Multi-Scale Control Signal-Aware Transformer for Motion Synthesis without Phase

no code implementations3 Mar 2023 Lintao Wang, Kun Hu, Lei Bai, Yu Ding, Wanli Ouyang, Zhiyong Wang

As past poses often contain useful auxiliary hints, in this paper, we propose a task-agnostic deep learning method, namely Multi-scale Control Signal-aware Transformer (MCS-T), with an attention based encoder-decoder architecture to discover the auxiliary information implicitly for synthesizing controllable motion without explicitly requiring auxiliary information such as phase.

Feature Engineering Motion Synthesis

Saliency Guided Contrastive Learning on Scene Images

no code implementations22 Feb 2023 Meilin Chen, Yizhou Wang, Shixiang Tang, Feng Zhu, Haiyang Yang, Lei Bai, Rui Zhao, Donglian Qi, Wanli Ouyang

Despite being feasible, recent works largely overlooked discovering the most discriminative regions for contrastive learning to object representations in scene images.

Contrastive Learning Representation Learning +1

Learning from pseudo-labels: deep networks improve consistency in longitudinal brain volume estimation

no code implementations8 Feb 2023 Geng Zhan, Dongang Wang, Mariano Cabezas, Lei Bai, Kain Kyle, Wanli Ouyang, Michael Barnett, Chenyu Wang

An accurate and robust quantitative measurement of brain volume change is paramount for translational research and clinical applications.

Fast-BEV: A Fast and Strong Bird's-Eye View Perception Baseline

1 code implementation29 Jan 2023 Yangguang Li, Bin Huang, Zeren Chen, Yufeng Cui, Feng Liang, Mingzhu Shen, Fenggang Liu, Enze Xie, Lu Sheng, Wanli Ouyang, Jing Shao

Our Fast-BEV consists of five parts, We novelly propose (1) a lightweight deployment-friendly view transformation which fast transfers 2D image feature to 3D voxel space, (2) an multi-scale image encoder which leverages multi-scale information for better performance, (3) an efficient BEV encoder which is particularly designed to speed up on-vehicle inference.

Data Augmentation

$β$-DARTS++: Bi-level Regularization for Proxy-robust Differentiable Architecture Search

1 code implementation16 Jan 2023 Peng Ye, Tong He, Baopu Li, Tao Chen, Lei Bai, Wanli Ouyang

To address the robustness problem, we first benchmark different NAS methods under a wide range of proxy data, proxy channels, proxy layers and proxy epochs, since the robustness of NAS under different kinds of proxies has not been explored before.

Neural Architecture Search

Crossing the Gap: Domain Generalization for Image Captioning

no code implementations CVPR 2023 Yuchen Ren, Zhendong Mao, Shancheng Fang, Yan Lu, Tong He, Hao Du, Yongdong Zhang, Wanli Ouyang

In this paper, we introduce a new setting called Domain Generalization for Image Captioning (DGIC), where the data from the target domain is unseen in the learning process.

Domain Generalization Image Captioning +1

Learning Multi-Modal Class-Specific Tokens for Weakly Supervised Dense Object Localization

no code implementations CVPR 2023 Lian Xu, Wanli Ouyang, Mohammed Bennamoun, Farid Boussaid, Dan Xu

Weakly supervised dense object localization (WSDOL) relies generally on Class Activation Mapping (CAM), which exploits the correlation between the class weights of the image classifier and the pixel-level features.

Object Localization Representation Learning +2

Open-Set Fine-Grained Retrieval via Prompting Vision-Language Evaluator

no code implementations CVPR 2023 Shijie Wang, Jianlong Chang, Haojie Li, Zhihui Wang, Wanli Ouyang, Qi Tian

PLEor could leverage pre-trained CLIP model to infer the discrepancies encompassing both pre-defined and unknown subcategories, called category-specific discrepancies, and transfer them to the backbone network trained in the close-set scenarios.

Knowledge Distillation Retrieval +1

Cap4Video: What Can Auxiliary Captions Do for Text-Video Retrieval?

3 code implementations CVPR 2023 Wenhao Wu, Haipeng Luo, Bo Fang, Jingdong Wang, Wanli Ouyang

Most existing text-video retrieval methods focus on cross-modal matching between the visual content of videos and textual query sentences.

Data Augmentation Retrieval +2

Ponder: Point Cloud Pre-training via Neural Rendering

no code implementations31 Dec 2022 Di Huang, Sida Peng, Tong He, Xiaowei Zhou, Wanli Ouyang

We propose a novel approach to self-supervised learning of point cloud representations by differentiable neural rendering.

3D Reconstruction Image Generation +2

Bidirectional Cross-Modal Knowledge Exploration for Video Recognition with Pre-trained Vision-Language Models

3 code implementations CVPR 2023 Wenhao Wu, Xiaohan Wang, Haipeng Luo, Jingdong Wang, Yi Yang, Wanli Ouyang

In this paper, we propose a novel framework called BIKE, which utilizes the cross-modal bridge to explore bidirectional knowledge: i) We introduce the Video Attribute Association mechanism, which leverages the Video-to-Text knowledge to generate textual auxiliary attributes for complementing video recognition.

Action Classification Action Recognition +2

MM-3DScene: 3D Scene Understanding by Customizing Masked Modeling with Informative-Preserved Reconstruction and Self-Distilled Consistency

no code implementations CVPR 2023 Mingye Xu, Mutian Xu, Tong He, Wanli Ouyang, Yali Wang, Xiaoguang Han, Yu Qiao

Besides, such scenes with progressive masking ratios can also serve to self-distill their intrinsic spatial consistency, requiring to learn the consistent representations from unmasked areas.

object-detection Object Detection +2

Frozen CLIP Model is An Efficient Point Cloud Backbone

1 code implementation8 Dec 2022 Xiaoshui Huang, Sheng Li, Wentao Qu, Tong He, Yifan Zuo, Wanli Ouyang

This paper introduces Efficient Point Cloud Learning (EPCL), an effective and efficient point cloud learner for directly training high-quality point cloud models with a frozen CLIP model.

Few-Shot Learning Semantic Segmentation

GD-MAE: Generative Decoder for MAE Pre-training on LiDAR Point Clouds

1 code implementation CVPR 2023 Honghui Yang, Tong He, Jiaheng Liu, Hua Chen, Boxi Wu, Binbin Lin, Xiaofei He, Wanli Ouyang

In contrast to previous 3D MAE frameworks, which either design a complex decoder to infer masked information from maintained regions or adopt sophisticated masking strategies, we instead propose a much simpler paradigm.

Reconstructing Hand-Held Objects from Monocular Video

no code implementations30 Nov 2022 Di Huang, Xiaopeng Ji, Xingyi He, Jiaming Sun, Tong He, Qing Shuai, Wanli Ouyang, Xiaowei Zhou

The key idea is that the hand motion naturally provides multiple views of the object and the motion can be reliably estimated by a hand pose tracker.

Hand Pose Estimation

ACE: Cooperative Multi-agent Q-learning with Bidirectional Action-Dependency

1 code implementation29 Nov 2022 Chuming Li, Jie Liu, Yinmin Zhang, Yuhong Wei, Yazhe Niu, Yaodong Yang, Yu Liu, Wanli Ouyang

In the learning phase, each agent minimizes the TD error that is dependent on how the subsequent agents have reacted to their chosen action.

Decision Making Q-Learning +2

3D-QueryIS: A Query-based Framework for 3D Instance Segmentation

no code implementations17 Nov 2022 Jiaheng Liu, Tong He, Honghui Yang, Rui Su, Jiayi Tian, Junran Wu, Hongcheng Guo, Ke Xu, Wanli Ouyang

Previous top-performing methods for 3D instance segmentation often maintain inter-task dependencies and the tendency towards a lack of robustness.

3D Instance Segmentation Semantic Segmentation

Boosting Semi-Supervised 3D Object Detection with Semi-Sampling

no code implementations14 Nov 2022 Xiaopei Wu, Yang Zhao, Liang Peng, Hua Chen, Xiaoshui Huang, Binbin Lin, Haifeng Liu, Deng Cai, Wanli Ouyang

When training a teacher-student semi-supervised framework, we randomly select gt samples and pseudo samples to both labeled frames and unlabeled frames, making a strong data augmentation for them.

3D Object Detection Data Augmentation +1

Stimulative Training of Residual Networks: A Social Psychology Perspective of Loafing

1 code implementation9 Oct 2022 Peng Ye, Shengji Tang, Baopu Li, Tao Chen, Wanli Ouyang

In this work, we aim to re-investigate the training process of residual networks from a novel social psychology perspective of loafing, and further propose a new training strategy to strengthen the performance of residual networks.

CLIP2Point: Transfer CLIP to Point Cloud Classification with Image-Depth Pre-training

1 code implementation3 Oct 2022 Tianyu Huang, Bowen Dong, Yunhan Yang, Xiaoshui Huang, Rynson W. H. Lau, Wanli Ouyang, WangMeng Zuo

To address this issue, we propose CLIP2Point, an image-depth pre-training method by contrastive learning to transfer CLIP to the 3D domain, and adapt it to point cloud classification.

Contrastive Learning Few-Shot Learning +3

Towards Frame Rate Agnostic Multi-Object Tracking

1 code implementation23 Sep 2022 Weitao Feng, Lei Bai, Yongqiang Yao, Fengwei Yu, Wanli Ouyang

In this paper, we propose a Frame Rate Agnostic MOT framework with a Periodic training Scheme (FAPS) to tackle the FraMOT problem for the first time.

Multi-Object Tracking

ZoomNAS: Searching for Whole-body Human Pose Estimation in the Wild

1 code implementation23 Aug 2022 Lumin Xu, Sheng Jin, Wentao Liu, Chen Qian, Wanli Ouyang, Ping Luo, Xiaogang Wang

We propose a single-network approach, termed ZoomNet, to take into account the hierarchical structure of the full human body and solve the scale variation of different body parts.

2D Human Pose Estimation Neural Architecture Search +1

Fine-grained Retrieval Prompt Tuning

no code implementations29 Jul 2022 Shijie Wang, Jianlong Chang, Zhihui Wang, Haojie Li, Wanli Ouyang, Qi Tian

In this paper, we develop Fine-grained Retrieval Prompt Tuning (FRPT), which steers a frozen pre-trained model to perform the fine-grained retrieval task from the perspectives of sample prompting and feature adaptation.


3D Interacting Hand Pose Estimation by Hand De-occlusion and Removal

1 code implementation22 Jul 2022 Hao Meng, Sheng Jin, Wentao Liu, Chen Qian, Mengxiang Lin, Wanli Ouyang, Ping Luo

Unlike most previous works that directly predict the 3D poses of two interacting hands simultaneously, we propose to decompose the challenging interacting hand pose estimation task and estimate the pose of each hand separately.

3D Interacting Hand Pose Estimation Hand Pose Estimation

Pose for Everything: Towards Category-Agnostic Pose Estimation

1 code implementation21 Jul 2022 Lumin Xu, Sheng Jin, Wang Zeng, Wentao Liu, Chen Qian, Wanli Ouyang, Ping Luo, Xiaogang Wang

In this paper, we introduce the task of Category-Agnostic Pose Estimation (CAPE), which aims to create a pose estimation model capable of detecting the pose of any class of object given only a few samples with keypoint definition.

2D Pose Estimation Pose Estimation

NSNet: Non-saliency Suppression Sampler for Efficient Video Recognition

no code implementations21 Jul 2022 Boyang xia, Wenhao Wu, Haoran Wang, Rui Su, Dongliang He, Haosen Yang, Xiaoran Fan, Wanli Ouyang

On the video level, a temporal attention module is learned under dual video-level supervisions on both the salient and the non-salient representations.

Action Recognition Video Classification +1

Action Recognition With Motion Diversification and Dynamic Selection

no code implementations TIP 2022 Peiqin Zhuang, Yu Guo, Zhipeng Yu, Luping Zhou, Lei Bai, Ding Liang, Zhiyong Wang, Yali Wang, Wanli Ouyang

To address this issue, we introduce a Motion Diversification and Selection (MoDS) module to generate diversified spatio-temporal motion features and then select the suitable motion representation dynamically for categorizing the input video.

Action Recognition

TransVG++: End-to-End Visual Grounding with Language Conditioned Vision Transformer

1 code implementation14 Jun 2022 Jiajun Deng, Zhengyuan Yang, Daqing Liu, Tianlang Chen, Wengang Zhou, Yanyong Zhang, Houqiang Li, Wanli Ouyang

For another, we devise Language Conditioned Vision Transformer that removes external fusion modules and reuses the uni-modal ViT for vision-language fusion at the intermediate layers.

Visual Grounding

Better Teacher Better Student: Dynamic Prior Knowledge for Knowledge Distillation

no code implementations13 Jun 2022 Zengyu Qiu, Xinzhu Ma, Kunlin Yang, Chunya Liu, Jun Hou, Shuai Yi, Wanli Ouyang

Besides, our DPK makes the performance of the student model positively correlated with that of the teacher model, which means that we can further boost the accuracy of students by applying larger teachers.

Image Classification Knowledge Distillation +3

Domain Invariant Masked Autoencoders for Self-supervised Learning from Multi-domains

no code implementations10 May 2022 Haiyang Yang, Meilin Chen, Yizhou Wang, Shixiang Tang, Feng Zhu, Lei Bai, Rui Zhao, Wanli Ouyang

While recent self-supervised learning methods have achieved good performances with evaluation set on the same domain as the training set, they will have an undesirable performance decrease when tested on a different domain.

Self-Supervised Learning

Unsupervised Learning of Accurate Siamese Tracking

1 code implementation CVPR 2022 Qiuhong Shen, Lei Qiao, Jinyang Guo, Peixia Li, Xin Li, Bo Li, Weitao Feng, Weihao Gan, Wei Wu, Wanli Ouyang

As unlimited self-supervision signals can be obtained by tracking a video along a cycle in time, we investigate evolving a Siamese tracker by tracking videos forward-backward.

Visual Object Tracking

SeCo: Separating Unknown Musical Visual Sounds with Consistency Guidance

no code implementations25 Mar 2022 Xinchi Zhou, Dongzhan Zhou, Wanli Ouyang, Hang Zhou, Ziwei Liu, Di Hu

Recent years have witnessed the success of deep learning on the visual sound separation task.

DR.VIC: Decomposition and Reasoning for Video Individual Counting

2 code implementations CVPR 2022 Tao Han, Lei Bai, Junyu Gao, Qi Wang, Wanli Ouyang

Instead of relying on the Multiple Object Tracking (MOT) techniques, we propose to solve the problem by decomposing all pedestrians into the initial pedestrians who existed in the first frame and the new pedestrians with separate identities in each following frame.

Crowd Counting Density Estimation +2

Backbone is All Your Need: A Simplified Architecture for Visual Object Tracking

1 code implementation10 Mar 2022 BoYu Chen, Peixia Li, Lei Bai, Lei Qiao, Qiuhong Shen, Bo Li, Weihao Gan, Wei Wu, Wanli Ouyang

Exploiting a general-purpose neural architecture to replace hand-wired designs or inductive biases has recently drawn extensive interest.

Visual Object Tracking

Multi-class Token Transformer for Weakly Supervised Semantic Segmentation

1 code implementation CVPR 2022 Lian Xu, Wanli Ouyang, Mohammed Bennamoun, Farid Boussaid, Dan Xu

To this end, we propose a Multi-class Token Transformer, termed as MCTformer, which uses multiple class tokens to learn interactions between the class tokens and the patch tokens.

Object Localization Weakly supervised Semantic Segmentation +1

$β$-DARTS: Beta-Decay Regularization for Differentiable Architecture Search

1 code implementation3 Mar 2022 Peng Ye, Baopu Li, Yikang Li, Tao Chen, Jiayuan Fan, Wanli Ouyang

Neural Architecture Search~(NAS) has attracted increasingly more attention in recent years because of its capability to design deep neural networks automatically.

Neural Architecture Search

3D Object Detection from Images for Autonomous Driving: A Survey

1 code implementation7 Feb 2022 Xinzhu Ma, Wanli Ouyang, Andrea Simonelli, Elisa Ricci

3D object detection from images, one of the fundamental and challenging problems in autonomous driving, has received increasing attention from both industry and academia in recent years.

3D Object Detection Autonomous Driving +1

Trajectory Forecasting from Detection with Uncertainty-Aware Motion Encoding

no code implementations3 Feb 2022 Pu Zhang, Lei Bai, Jianru Xue, Jianwu Fang, Nanning Zheng, Wanli Ouyang

Trajectories obtained from object detection and tracking are inevitably noisy, which could cause serious forecasting errors to predictors built on ground truth trajectories.

object-detection Object Detection +1

Pseudo-Labeled Auto-Curriculum Learning for Semi-Supervised Keypoint Localization

no code implementations ICLR 2022 Can Wang, Sheng Jin, Yingda Guan, Wentao Liu, Chen Qian, Ping Luo, Wanli Ouyang

PL approaches apply pseudo-labels to unlabeled data, and then train the model with a combination of the labeled and pseudo-labeled data iteratively.

RePre: Improving Self-Supervised Vision Transformer with Reconstructive Pre-training

no code implementations18 Jan 2022 Luya Wang, Feng Liang, Yangguang Li, Honggang Zhang, Wanli Ouyang, Jing Shao

Recently, self-supervised vision transformers have attracted unprecedented attention for their impressive representation learning ability.

Contrastive Learning Representation Learning

Accelerating Neural Network Optimization Through an Automated Control Theory Lens

no code implementations CVPR 2022 Jiahao Wang, Baoyuan Wu, Rui Su, Mingdeng Cao, Shuwei Shi, Wanli Ouyang, Yujiu Yang

We conduct experiments both from a control theory lens through a phase locus verification and from a network training lens on several models, including CNNs, Transformers, MLPs, and on benchmark datasets.

b-DARTS: Beta-Decay Regularization for Differentiable Architecture Search

1 code implementation CVPR 2022 Peng Ye, Baopu Li, Yikang Li, Tao Chen, Jiayuan Fan, Wanli Ouyang

Neural Architecture Search (NAS) has attracted increasingly more attention in recent years because of its capability to design deep neural network automatically.

Neural Architecture Search

A Continuous Mapping For Augmentation Design

no code implementations NeurIPS 2021 Keyu Tian, Chen Lin, Ser Nam Lim, Wanli Ouyang, Puneet Dokania, Philip Torr

Automated data augmentation (ADA) techniques have played an important role in boosting the performance of deep models.

Data Augmentation

Unsupervised Contrastive Learning with Simple Transformation for 3D Point Cloud Data

no code implementations13 Oct 2021 Jincen Jiang, Xuequan Lu, Wanli Ouyang, Meili Wang

Though a number of point cloud learning methods have been proposed to handle unordered points, most of them are supervised and require labels for training.

3D Object Classification Classification +3

Supervision Exists Everywhere: A Data Efficient Contrastive Language-Image Pre-training Paradigm

3 code implementations ICLR 2022 Yangguang Li, Feng Liang, Lichen Zhao, Yufeng Cui, Wanli Ouyang, Jing Shao, Fengwei Yu, Junjie Yan

Recently, large-scale Contrastive Language-Image Pre-training (CLIP) has attracted unprecedented attention for its impressive zero-shot recognition ability and excellent transferability to downstream tasks.

Zero-Shot Learning

Deep Instance Segmentation with Automotive Radar Detection Points

no code implementations5 Oct 2021 Jianan Liu, Weiyi Xiong, Liping Bai, Yuxuan Xia, Tao Huang, Wanli Ouyang, Bing Zhu

Automotive radar provides reliable environmental perception in all-weather conditions with affordable cost, but it hardly supplies semantic and geometry information due to the sparsity of radar detection points.

Autonomous Driving Clustering +2

Towards Balanced Learning for Instance Recognition

no code implementations23 Aug 2021 Jiangmiao Pang, Kai Chen, Qi Li, Zhihai Xu, Huajun Feng, Jianping Shi, Wanli Ouyang, Dahua Lin

In this work, we carefully revisit the standard training practice of detectors, and find that the detection performance is often limited by the imbalance during the training process, which generally consists in three levels - sample level, feature level, and objective level.

BN-NAS: Neural Architecture Search with Batch Normalization

1 code implementation ICCV 2021 BoYu Chen, Peixia Li, Baopu Li, Chen Lin, Chuming Li, Ming Sun, Junjie Yan, Wanli Ouyang

We present BN-NAS, neural architecture search with Batch Normalization (BN-NAS), to accelerate neural architecture search (NAS).

Neural Architecture Search

PSViT: Better Vision Transformer via Token Pooling and Attention Sharing

no code implementations7 Aug 2021 BoYu Chen, Peixia Li, Baopu Li, Chuming Li, Lei Bai, Chen Lin, Ming Sun, Junjie Yan, Wanli Ouyang

Then, a compact set of the possible combinations for different token pooling and attention sharing mechanisms are constructed.

Leveraging Auxiliary Tasks with Affinity Learning for Weakly Supervised Semantic Segmentation

1 code implementation ICCV 2021 Lian Xu, Wanli Ouyang, Mohammed Bennamoun, Farid Boussaid, Ferdous Sohel, Dan Xu

Motivated by the significant inter-task correlation, we propose a novel weakly supervised multi-task framework termed as AuxSegNet, to leverage saliency detection and multi-label image classification as auxiliary tasks to improve the primary task of semantic segmentation using only image-level ground-truth labels.

Auxiliary Learning Multi-Label Image Classification +5

Mutual CRF-GNN for Few-Shot Learning

no code implementations CVPR 2021 Shixiang Tang, Dapeng Chen, Lei Bai, Kaijian Liu, Yixiao Ge, Wanli Ouyang

In this MCGN, the labels and features of support data are used by the CRF for inferring GNN affinities in a principled and probabilistic way.

Few-Shot Learning

Layerwise Optimization by Gradient Decomposition for Continual Learning

no code implementations CVPR 2021 Shixiang Tang, Dapeng Chen, Jinguo Zhu, Shijie Yu, Wanli Ouyang

The gradient for update should be close to the gradient of the new task, consistent with the gradients shared by all old tasks, and orthogonal to the space spanned by the gradients specific to the old tasks.

Continual Learning

Delving into Localization Errors for Monocular 3D Object Detection

1 code implementation CVPR 2021 Xinzhu Ma, Yinmin Zhang, Dan Xu, Dongzhan Zhou, Shuai Yi, Haojie Li, Wanli Ouyang

Estimating 3D bounding boxes from monocular images is an essential component in autonomous driving, while accurate 3D object detection from this kind of data is very challenging.

Autonomous Driving Monocular 3D Object Detection +1

Gradient Regularized Contrastive Learning for Continual Domain Adaptation

no code implementations23 Mar 2021 Shixiang Tang, Peng Su, Dapeng Chen, Wanli Ouyang

To better understand this issue, we study the problem of continual domain adaptation, where the model is presented with a labelled source domain and a sequence of unlabelled target domains.

Contrastive Learning Domain Adaptation

Probabilistic Graph Attention Network with Conditional Kernels for Pixel-Wise Prediction

no code implementations8 Jan 2021 Dan Xu, Xavier Alameda-Pineda, Wanli Ouyang, Elisa Ricci, Xiaogang Wang, Nicu Sebe

In contrast to previous works directly considering multi-scale feature maps obtained from the inner layers of a primary CNN architecture, and simply fusing the features with weighted averaging or concatenation, we propose a probabilistic graph attention network structure based on a novel Attention-Gated Conditional Random Fields (AG-CRFs) model for learning and fusing multi-scale representations in a principled manner.

Graph Attention Monocular Depth Estimation +1

Inception Convolution with Efficient Dilation Search

1 code implementation CVPR 2021 Jie Liu, Chuming Li, Feng Liang, Chen Lin, Ming Sun, Junjie Yan, Wanli Ouyang, Dong Xu

To develop a practical method for learning complex inception convolution based on the data, a simple but effective search algorithm, referred to as efficient dilation optimization (EDO), is developed.

Human Detection Instance Segmentation +4

DETR for Crowd Pedestrian Detection

1 code implementation12 Dec 2020 Matthieu Lin, Chuming Li, Xingyuan Bu, Ming Sun, Chen Lin, Junjie Yan, Wanli Ouyang, Zhidong Deng

Furthermore, the bipartite match of ED harms the training efficiency due to the large ground truth number in crowd scenes.

Pedestrian Detection

Full Matching on Low Resolution for Disparity Estimation

no code implementations10 Dec 2020 Hong Zhang, Shenglun Chen, Zhihui Wang, Haojie Li, Wanli Ouyang

To this end, we first propose to decompose the full matching task into multiple stages of the cost aggregation module.

Disparity Estimation

Direct Depth Learning Network for Stereo Matching

no code implementations10 Dec 2020 Hong Zhang, Haojie Li, Shenglun Chen, Tiantian Yan, Zhihui Wang, Guo Lu, Wanli Ouyang

To make the Adaptive-Grained Depth Refinement stage robust to the coarse depth and adaptive to the depth range of the points, the Granularity Uncertainty is introduced to Adaptive-Grained Depth Refinement stage.

Autonomous Driving Depth Estimation +1

Temporal-Channel Transformer for 3D Lidar-Based Video Object Detection in Autonomous Driving

no code implementations27 Nov 2020 Zhenxun Yuan, Xiao Song, Lei Bai, Wengang Zhou, Zhe Wang, Wanli Ouyang

As a special design of this transformer, the information encoded in the encoder is different from that in the decoder, i. e. the encoder encodes temporal-channel information of multiple frames while the decoder decodes the spatial-channel information for the current frame in a voxel-wise manner.

3D Object Detection Autonomous Driving +2

Adaptive Gradient Method with Resilience and Momentum

no code implementations21 Oct 2020 Jie Liu, Chen Lin, Chuming Li, Lu Sheng, Ming Sun, Junjie Yan, Wanli Ouyang

Several variants of stochastic gradient descent (SGD) have been proposed to improve the learning effectiveness and efficiency when training deep neural networks, among which some recent influential attempts would like to adaptively control the parameter-wise learning rate (e. g., Adam and RMSProp).

Category-specific Semantic Coherency Learning for Fine-grained Image Recognition

no code implementations12 Oct 2020 Shijie Wang, Zhihui Wang, Haojie Li, Wanli Ouyang

Existing deep learning based weakly supervised fine-grained image recognition (WFGIR) methods usually pick out the discriminative regions from the high-level feature (HLF) maps directly.

Fine-Grained Image Recognition

Improving Auto-Augment via Augmentation-Wise Weight Sharing

1 code implementation NeurIPS 2020 Keyu Tian, Chen Lin, Ming Sun, Luping Zhou, Junjie Yan, Wanli Ouyang

On CIFAR-10, we achieve a top-1 error rate of 1. 24%, which is currently the best performing single model without extra training data.

SAMOT: Switcher-Aware Multi-Object Tracking and Still Another MOT Measure

no code implementations22 Sep 2020 Weitao Feng, Zhihao Hu, Baopu Li, Weihao Gan, Wei Wu, Wanli Ouyang

Besides, we propose a new MOT evaluation measure, Still Another IDF score (SAIDF), aiming to focus more on identity issues. This new measure may overcome some problems of the previous measures and provide a better insight for identity issues in MOT.

Multi-Object Tracking

Improving Deep Video Compression by Resolution-adaptive Flow Coding

no code implementations ECCV 2020 Zhihao Hu, Zhenghao Chen, Dong Xu, Guo Lu, Wanli Ouyang, Shuhang Gu

In this work, we propose a new framework called Resolution-adaptive Flow Coding (RaFC) to effectively compress the flow maps globally and locally, in which we use multi-resolution representations instead of single-resolution representations for both the input flow maps and the output motion features of the MV encoder.

Optical Flow Estimation Video Compression

Exploring the Hierarchy in Relation Labels for Scene Graph Generation

no code implementations12 Sep 2020 Yi Zhou, Shuyang Sun, Chao Zhang, Yikang Li, Wanli Ouyang

By assigning each relationship a single label, current approaches formulate the relationship detection as a classification problem.

Graph Generation Relationship Detection +1

Rethinking Pseudo-LiDAR Representation

1 code implementation ECCV 2020 Xinzhu Ma, Shinan Liu, Zhiyi Xia, Hongwen Zhang, Xingyu Zeng, Wanli Ouyang

Based on this observation, we design an image based CNN detector named Patch-Net, which is more generalized and can be instantiated as pseudo-LiDAR based 3D detectors.

Whole-Body Human Pose Estimation in the Wild

2 code implementations ECCV 2020 Sheng Jin, Lumin Xu, Jin Xu, Can Wang, Wentao Liu, Chen Qian, Wanli Ouyang, Ping Luo

This paper investigates the task of 2D human whole-body pose estimation, which aims to localize dense landmarks on the entire human body including face, hands, body, and feet.

2D Human Pose Estimation Facial Landmark Detection +2

Differentiable Hierarchical Graph Grouping for Multi-Person Pose Estimation

no code implementations ECCV 2020 Sheng Jin, Wentao Liu, Enze Xie, Wenhai Wang, Chen Qian, Wanli Ouyang, Ping Luo

The modules of HGG can be trained end-to-end with the keypoint detection network and is able to supervise the grouping process in a hierarchical manner.

2D Human Pose Estimation Clustering +4

3D Human Mesh Regression with Dense Correspondence

3 code implementations CVPR 2020 Wang Zeng, Wanli Ouyang, Ping Luo, Wentao Liu, Xiaogang Wang

This paper proposes a model-free 3D human mesh estimation framework, named DecoMR, which explicitly establishes the dense correspondence between the mesh and the local image features in the UV space (i. e. a 2D space used for texture mapping of 3D mesh).

3D Human Pose Estimation 3D Human Reconstruction +1

Scope Head for Accurate Localization in Object Detection

no code implementations11 May 2020 Geng Zhan, Dan Xu, Guo Lu, Wei Wu, Chunhua Shen, Wanli Ouyang

Existing anchor-based and anchor-free object detectors in multi-stage or one-stage pipelines have achieved very promising detection performance.

object-detection Object Detection +1

Content Adaptive and Error Propagation Aware Deep Video Compression

no code implementations ECCV 2020 Guo Lu, Chunlei Cai, Xiaoyun Zhang, Li Chen, Wanli Ouyang, Dong Xu, Zhiyong Gao

Therefore, the encoder is adaptive to different video contents and achieves better compression performance by reducing the domain gap between the training and testing datasets.

Video Compression

Channel Pruning Guided by Classification Loss and Feature Importance

no code implementations15 Mar 2020 Jinyang Guo, Wanli Ouyang, Dong Xu

To this end, we propose a new strategy to suppress the influence of unimportant features (i. e., the features will be removed at the next pruning stage).

Classification Feature Importance +1

Equalization Loss for Long-Tailed Object Recognition

1 code implementation CVPR 2020 Jingru Tan, Changbao Wang, Buyu Li, Quanquan Li, Wanli Ouyang, Changqing Yin, Junjie Yan

Based on it, we propose a simple but effective loss, named equalization loss, to tackle the problem of long-tailed rare categories by simply ignoring those gradients for rare categories.

Long-tail Learning object-detection +2

EcoNAS: Finding Proxies for Economical Neural Architecture Search

no code implementations CVPR 2020 Dongzhan Zhou, Xinchi Zhou, Wenwei Zhang, Chen Change Loy, Shuai Yi, Xuesen Zhang, Wanli Ouyang

While many methods have been proposed to improve the efficiency of NAS, the search progress is still laborious because training and evaluating plausible architectures over large search space is time-consuming.

Neural Architecture Search

Learning 3D Human Shape and Pose from Dense Body Parts

1 code implementation31 Dec 2019 Hongwen Zhang, Jie Cao, Guo Lu, Wanli Ouyang, Zhenan Sun

Reconstructing 3D human shape and pose from monocular images is challenging despite the promising results achieved by the most recent learning-based methods.

Ranked #57 on 3D Human Pose Estimation on 3DPW (MPJPE metric)

3D human pose and shape estimation 3D Human Reconstruction +2

Computation Reallocation for Object Detection

no code implementations ICLR 2020 Feng Liang, Chen Lin, Ronghao Guo, Ming Sun, Wei Wu, Junjie Yan, Wanli Ouyang

However, classification allocation pattern is usually adopted directly to object detector, which is proved to be sub-optimal.

Instance Segmentation Neural Architecture Search +3

A Shape Transformation-based Dataset Augmentation Framework for Pedestrian Detection

no code implementations15 Dec 2019 Zhe Chen, Wanli Ouyang, Tongliang Liu, DaCheng Tao

Alternatively, to access much more natural-looking pedestrians, we propose to augment pedestrian detection datasets by transforming real pedestrians from the same dataset into different shapes.

Pedestrian Detection

TRB: A Novel Triplet Representation for Understanding 2D Human Body

2 code implementations ICCV 2019 Haodong Duan, Kwan-Yee Lin, Sheng Jin, Wentao Liu, Chen Qian, Wanli Ouyang

In this paper, we propose the Triplet Representation for Body (TRB) -- a compact 2D human body representation, with skeleton keypoints capturing human pose information and contour keypoints containing human shape information.

Conditional Image Generation Open-Ended Question Answering

Improving One-shot NAS by Suppressing the Posterior Fading

no code implementations CVPR 2020 Xiang Li, Chen Lin, Chuming Li, Ming Sun, Wei Wu, Junjie Yan, Wanli Ouyang

In this paper, we analyse existing weight sharing one-shot NAS approaches from a Bayesian point of view and identify the posterior fading problem, which compromises the effectiveness of shared weights.

Neural Architecture Search object-detection +2

IntersectGAN: Learning Domain Intersection for Generating Images with Multiple Attributes

no code implementations21 Sep 2019 Zehui Yao, Boyan Zhang, Zhiyong Wang, Wanli Ouyang, Dong Xu, Dagan Feng

For example, given two image domains $X_1$ and $X_2$ with certain attributes, the intersection $X_1 \cap X_2$ denotes a new domain where images possess the attributes from both $X_1$ and $X_2$ domains.

GradNet: Gradient-Guided Network for Visual Object Tracking

2 code implementations ICCV 2019 Peixia Li, Bo-Yu Chen, Wanli Ouyang, Dong Wang, Xiaoyun Yang, Huchuan Lu

In this work, we propose a novel gradient-guided network to exploit the discriminative information in gradients and update the template in the siamese network through feed-forward and backward operations.

Ranked #3 on Visual Object Tracking on OTB-2015 (Precision metric)

Template Matching Visual Object Tracking +1

Structured Modeling of Joint Deep Feature and Prediction Refinement for Salient Object Detection

1 code implementation ICCV 2019 Yingyue Xu, Dan Xu, Xiaopeng Hong, Wanli Ouyang, Rongrong Ji, Min Xu, Guoying Zhao

We formulate the CRF graphical model that involves message-passing of feature-feature, feature-prediction, and prediction-prediction, from the coarse scale to the finer scale, to update the features and the corresponding predictions.

object-detection RGB Salient Object Detection +1

Crowd Counting with Deep Structured Scale Integration Network

no code implementations ICCV 2019 Lingbo Liu, Zhilin Qiu, Guanbin Li, Shufan Liu, Wanli Ouyang, Liang Lin

Automatic estimation of the number of people in unconstrained crowded scenes is a challenging task and one major difficulty stems from the huge scale variation of people.

Crowd Counting Representation Learning

Improving Action Localization by Progressive Cross-stream Cooperation

no code implementations CVPR 2019 Rui Su, Wanli Ouyang, Luping Zhou, Dong Xu

Specifically, we first generate a larger set of region proposals by combining the latest region proposals from both streams, from which we can readily obtain a larger set of labelled training samples to help learn better action detection models.

Action Classification Action Detection +2

AM-LFS: AutoML for Loss Function Search

1 code implementation ICCV 2019 Chuming Li, Yuan Xin, Chen Lin, Minghao Guo, Wei Wu, Wanli Ouyang, Junjie Yan

The key contribution of this work is the design of search space which can guarantee the generalization and transferability on different vision tasks by including a bunch of existing prevailing loss functions in a unified formulation.


Contextualized Spatial-Temporal Network for Taxi Origin-Destination Demand Prediction

no code implementations15 May 2019 Lingbo Liu, Zhilin Qiu, Guanbin Li, Qing Wang, Wanli Ouyang, Liang Lin

Finally, a GCC module is applied to model the correlation between all regions by computing a global correlation feature as a weighted sum of all regional features, with the weights being calculated as the similarity between the corresponding region pairs.

Libra R-CNN: Towards Balanced Learning for Object Detection

5 code implementations CVPR 2019 Jiangmiao Pang, Kai Chen, Jianping Shi, Huajun Feng, Wanli Ouyang, Dahua Lin

In this work, we carefully revisit the standard training practice of detectors, and find that the detection performance is often limited by the imbalance during the training process, which generally consists in three levels - sample level, feature level, and objective level.

object-detection Object Detection

Feature Intertwiner for Object Detection

2 code implementations ICLR 2019 Hongyang Li, Bo Dai, Shaoshuai Shi, Wanli Ouyang, Xiaogang Wang

We argue that the reliable set could guide the feature learning of the less reliable set during training - in spirit of student mimicking teacher behavior and thus pushing towards a more compact class centroid in the feature space.

object-detection Object Detection

Accurate Monocular Object Detection via Color-Embedded 3D Reconstruction for Autonomous Driving

no code implementations27 Mar 2019 Xinzhu Ma, Zhihui Wang, Haojie Li, Peng-Bo Zhang, Xin Fan, Wanli Ouyang

To this end, we first leverage a stand-alone module to transform the input data from 2D image plane to 3D point clouds space for a better input representation, then we perform the 3D detection using PointNet backbone net to obtain objects 3D locations, dimensions and orientations.

3D Reconstruction Autonomous Driving +2

SR-LSTM: State Refinement for LSTM towards Pedestrian Trajectory Prediction

1 code implementation CVPR 2019 Pu Zhang, Wanli Ouyang, Pengfei Zhang, Jianru Xue, Nanning Zheng

In order to address this issue, we propose a data-driven state refinement module for LSTM network (SR-LSTM), which activates the utilization of the current intention of neighbors, and jointly and iteratively refines the current states of all participants in the crowd through a message passing mechanism.

Pedestrian Trajectory Prediction Trajectory Prediction