Search

Your search keyword '"Wen, Yuqing"' showing total 214 results
Start Over Author "Wen, Yuqing"
214 results on '"Wen, Yuqing"'

1. Panacea+: Panoramic and Controllable Video Generation for Autonomous Driving

Author

Wen, Yuqing, Zhao, Yucheng, Liu, Yingfei, Huang, Binyuan, Jia, Fan, Wang, Yanhui, Zhang, Chi, Wang, Tiancai, Sun, Xiaoyan, and Zhang, Xiangyu

Subjects
Computer Science - Computer Vision and Pattern Recognition
Abstract

The field of autonomous driving increasingly demands high-quality annotated video training data. In this paper, we propose Panacea+, a powerful and universally applicable framework for generating video data in driving scenes. Built upon the foundation of our previous work, Panacea, Panacea+ adopts a multi-view appearance noise prior mechanism and a super-resolution module for enhanced consistency and increased resolution. Extensive experiments show that the generated video samples from Panacea+ greatly benefit a wide range of tasks on different datasets, including 3D object tracking, 3D object detection, and lane detection tasks on the nuScenes and Argoverse 2 dataset. These results strongly prove Panacea+ to be a valuable data generation framework for autonomous driving., Comment: Project page: https://panacea-ad.github.io/. arXiv admin note: text overlap with arXiv:2311.16813

Published
2024

2. SubjectDrive: Scaling Generative Data in Autonomous Driving via Subject Control

Author

Huang, Binyuan, Wen, Yuqing, Zhao, Yucheng, Hu, Yaosi, Liu, Yingfei, Jia, Fan, Mao, Weixin, Wang, Tiancai, Zhang, Chi, Chen, Chang Wen, Chen, Zhenzhong, and Zhang, Xiangyu

Subjects
Computer Science - Computer Vision and Pattern Recognition, Computer Science - Robotics
Abstract

Autonomous driving progress relies on large-scale annotated datasets. In this work, we explore the potential of generative models to produce vast quantities of freely-labeled data for autonomous driving applications and present SubjectDrive, the first model proven to scale generative data production in a way that could continuously improve autonomous driving applications. We investigate the impact of scaling up the quantity of generative data on the performance of downstream perception models and find that enhancing data diversity plays a crucial role in effectively scaling generative data production. Therefore, we have developed a novel model equipped with a subject control mechanism, which allows the generative model to leverage diverse external data sources for producing varied and useful data. Extensive evaluations confirm SubjectDrive's efficacy in generating scalable autonomous driving training data, marking a significant step toward revolutionizing data production methods in this field., Comment: Project page: https://subjectdrive.github.io/

Published
2024

3. ADriver-I: A General World Model for Autonomous Driving

Author

Jia, Fan, Mao, Weixin, Liu, Yingfei, Zhao, Yucheng, Wen, Yuqing, Zhang, Chi, Zhang, Xiangyu, and Wang, Tiancai

Subjects
Computer Science - Computer Vision and Pattern Recognition, Computer Science - Robotics
Abstract

Typically, autonomous driving adopts a modular design, which divides the full stack into perception, prediction, planning and control parts. Though interpretable, such modular design tends to introduce a substantial amount of redundancy. Recently, multimodal large language models (MLLM) and diffusion techniques have demonstrated their superior performance on comprehension and generation ability. In this paper, we first introduce the concept of interleaved vision-action pair, which unifies the format of visual features and control signals. Based on the vision-action pairs, we construct a general world model based on MLLM and diffusion model for autonomous driving, termed ADriver-I. It takes the vision-action pairs as inputs and autoregressively predicts the control signal of the current frame. The generated control signals together with the historical vision-action pairs are further conditioned to predict the future frames. With the predicted next frame, ADriver-I performs further control signal prediction. Such a process can be repeated infinite times, ADriver-I achieves autonomous driving in the world created by itself. Extensive experiments are conducted on nuScenes and our large-scale private datasets. ADriver-I shows impressive performance compared to several constructed baselines. We hope our ADriver-I can provide some new insights for future autonomous driving and embodied intelligence., Comment: Tech Report

Published
2023

4. Preparation and electrochemical properties of two-dimensional nanorod structure CoMn2O4/CC electrode materials

Author

LIU Jiahong, LU Shichen, WANG Likui, LIU Yanhong, WEN Yuqing, and SHANG Wei

Subjects
supercapacitor, manganese-based metal-organic framework derivative, electrode material, flexibility, Materials of engineering and construction. Mechanics of materials, TA401-492
Abstract

Metal oxides are often used as electrode materials for supercapacitors because of their high capacity， low cost， suitability for commercialization， and environmental friendliness. In this study， Mn-MOF was used as a precursor and placed in ethanol for ion exchange and etching reaction with Co（NO3）2， followed by heat treatment in air， and finally highly crystalline CoMn2O4 two-dimensional nanorod structures were obtained on a carbon cloth（CC） substrate. SEM and XRD analyses of CoMn2O4/CC prepared with different Co（NO3）2 additions and Mn2O3/CC obtained by direct heat treatment were performed， and electrochemical properties were measured by cyclic voltammetry test， constant current charge/discharge test and AC impedance test. The results show that the parent structure of CoMn2O4/CC is relatively well preserved after 0.3 g Co（NO3）2 etching， forming a nano-hollow structure， and the vertical nanorod structures grow in situ on carbon cloth after heat treatment wrapping carbon fibers uniformly and densely， ensuring high mechanical stability and electrical conductivity due to the absence of any binder addition. The electrode material has an area specific capacitance of 809.8 mF·cm-2 at the current density of 1.2 mA·cm-2 and the capacitance retention is 79.1% after 5000 cycles at the current density of 5 mA·cm-2，showing potential application prospects.

Published
2024
Full Text
View/download PDF

6. Corrosion resistance behavior of different MOF superhydrophobic coatings on magnesium alloy surface

Author

LIU Jinyu, ZHANG Zhiyuan, WANG Dong, JIANG Shiquan, WEN Yuqing, and SHANG Wei

Subjects
magnesium alloy, mof, superhydrophobic coating, hydrophobic stability, corrosion resistance behavior, Materials of engineering and construction. Mechanics of materials, TA401-492
Abstract

In order to study the corrosion resistance of three different superhydrophobic coatings (MZS-1, MZS-2 and ZnO@ZIF-8) on AZ91D magnesium alloy surface in 5% (mass fraction) NaCl solution. The microstructure, wettability and corrosion resistance of the superhydrophobic composite coating were tested and characterized by field emission scanning electron microscope, static contact angle tester, electrochemical workstation and salt spray tester, respectively. The results show that the corrosion of the superhydrophobic coatings does not occur until 192 h after salt spray treatment among the three types of superhydrophobic coatings, and the corrosion of the MZS-1 superhydrophobic coating is the most serious. The surface pitting of the MZS-2 superhydrophobic coating doesn't occur until 240 h later, and the contact angle is still high after salt spray treatment, so the corrosion resistance of the MZS-2 composite coating is the best.The polarization curve tests indicate that the corrosion current density of three superhydrophobic coatings are still one order of magnitude lower than that of the metal matrix after salt spray treatment for 240 h, showing excellent corrosion resistance. The superhydrophobic coating can effectively increase the corrosion resistance of metal materials.It can effectively prevent the infiltration of corrosive ions and provide long-term protection for the matrix because of its water repellency.

Published
2024
Full Text
View/download PDF

43. Panacea: Panoramic and Controllable Video Generation for Autonomous Driving

Author

Wen, Yuqing, Zhao, Yucheng, Liu, Yingfei, Jia, Fan, Wang, Yanhui, Luo, Chong, Zhang, Chi, Wang, Tiancai, Sun, Xiaoyan, Zhang, Xiangyu, Wen, Yuqing, Zhao, Yucheng, Liu, Yingfei, Jia, Fan, Wang, Yanhui, Luo, Chong, Zhang, Chi, Wang, Tiancai, Sun, Xiaoyan, and Zhang, Xiangyu

Abstract

The field of autonomous driving increasingly demands high-quality annotated training data. In this paper, we propose Panacea, an innovative approach to generate panoramic and controllable videos in driving scenarios, capable of yielding an unlimited numbers of diverse, annotated samples pivotal for autonomous driving advancements. Panacea addresses two critical challenges: 'Consistency' and 'Controllability.' Consistency ensures temporal and cross-view coherence, while Controllability ensures the alignment of generated content with corresponding annotations. Our approach integrates a novel 4D attention and a two-stage generation pipeline to maintain coherence, supplemented by the ControlNet framework for meticulous control by the Bird's-Eye-View (BEV) layouts. Extensive qualitative and quantitative evaluations of Panacea on the nuScenes dataset prove its effectiveness in generating high-quality multi-view driving-scene videos. This work notably propels the field of autonomous driving by effectively augmenting the training dataset used for advanced BEV perception techniques., Comment: Project page: https://panacea-ad.github.io

Published
2023

Catalog

Books, media, physical & digital resources
See catalog results