检索结果-内蒙古大学图书馆

Fast Instrument Learning with Faster Rates

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Wang, Ziyu Zhou, Yuhao Zhu, Jun Dept. of Comp. Sci. and Tech. BNRist Center State Key Lab for Intell. Tech. & Sys. Institute for AI Tsinghua-Bosch Joint Center for ML Tsinghua University China

关键词： Learning systems

Detection Transformer with Stable Matching

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Liu, Shilong Ren, Tianhe Chen, Jiayu Zeng, Zhaoyang Zhang, Hao Li, Feng Li, Hongyang Huang, Jun Su, Hang Zhu, Jun Zhang, Lei Dept. of Comp. Sci. and Tech. BNRist Center State Key Lab for Intell. Tech. & Sys. Institute for AI Tsinghua-Bosch Joint Center for ML Tsinghua University China Alibaba Group China The Hong Kong University of Science and Technology Hong Kong South China University of Technology China

This paper is concerned with the matching stability problem across different decoder layers in DEtection TRansformers (DETR). We point out that the unstable matching in DETR is caused by a multi-optimization path problem, which is highlighted by the one-to-one matching design in DETR. To address this problem, we show that the most important design is to use and only use positional metrics (like IOU) to supervise classification scores of positive examples. Under the principle, we propose two simple yet effective modifications by integrating positional metrics to DETR’s classification loss and matching cost, named position-supervised loss and position-modulated cost. We verify our methods on several DETR variants. Our methods show consistent improvements over baselines. By integrating our methods with DINO, we achieve 50.4 and 51.5 AP on the COCO detection benchmark using ResNet-50 backbones under 1× (12 epochs) and 2× (24 epochs) training settings, achieving a new record under the same setting. We achieve 63.8 AP on COCO detection test-dev with a Swin-Large backbone. Our code will be made available at https://***/IDEA-Research/Stable-DINO. Copyright © 2023, The Authors. All rights reserved.

关键词： Machine learning

One Transformer Fits All Distributions in Multi-Modal Diffusion at Scale

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Bao, Fan Nie, Shen Xue, Kaiwen Li, Chongxuan Pu, Shi Wang, Yaole Yue, Gang Cao, Yue Su, Hang Zhu, Jun Dept. of Comp. Sci. & Tech. Institute for AI Tsinghua-Huawei Joint Center for AI BNRist Center State Key Lab for Intell. Tech. & Sys. Tsinghua University China ShengShu Beijing China Gaoling School of AI Renmin University of China Beijing Key Lab of Big Data Management and Analysis Methods Beijing China Beijing Academy of Artificial Intelligence China Guangzhou China

This paper proposes a unified diffusion framework (dubbed UniDiffuser) to fit all distributions relevant to a set of multi-modal data in one model. Our key insight is – learning diffusion models for marginal, conditional, and joint distributions can be unified as predicting the noise in the perturbed data, where the perturbation levels (i.e. timesteps) can be different for different modalities. Inspired by the unified view, UniDiffuser learns all distributions simultaneously with a minimal modification to the original diffusion model – perturbs data in all modalities instead of a single modality, inputs individual timesteps in different modalities, and predicts the noise of all modalities instead of a single modality. UniDiffuser is parameterized by a transformer for diffusion models to handle input types of different modalities. Implemented on large-scale paired image-text data, UniDiffuser is able to perform image, text, text-to-image, image-to-text, and image-text pair generation by setting proper timesteps without additional overhead. In particular, UniDiffuser is able to produce perceptually realistic samples in all tasks and its quantitative results (e.g., the FID and CLIP score) are not only superior to existing general-purpose models but also comparable to the bespoken models (e.g., Stable Diffusion and DALL·E 2) in representative tasks (e.g., text-to-image generation). Our code is available at https://***/thu-ml/unidiffuser. Copyright © 2023, The Authors. All rights reserved.

关键词： Diffusion

The Winning Solution to the iFLYTEK Challenge 2021 Cultivated Land Extraction from High-Resolution Remote Sensing Images 14

学校读者我要写书评

暂无评论

The Winning Solution to the iFLYTEK Challenge 2021 Cultivate...

14th International Conference on Advanced Computational Intelligence

作者： Zhao, Zhen Liu, Yuqiu Zhang, Gang Tang, Liang Hu, Xiaolin Beijing Forestry Univ Sch Technol Beijing 1100091 Peoples R China Tsinghua Univ Dept Comp Sci & Technol State Key Lab Intelligent Tech & Sys BNRist Beijing 100084 Peoples R China Tsinghua Univ THBI Beijing 100084 Peoples R China Chinese Inst Brain Res CIBR Beijing 102206 Peoples R China Chongqing Univ Technol Key Lab Adv Mfg Technol Automobile Parts Minist Educ Chongqing 400054 Peoples R China

ISBN: (数字)9781665470452

ISBN: (纸本)9781665470469;9781665470452

Extracting cultivated land accurately from high-resolution remote images is a basic task for precision agriculture. This paper introduces our solution to iFLYTEK challenge 2021 cultivated land extraction from high-resolution remote sensing images. We established a highly effective and efficient pipeline to solve this problem. We first divided the original images into small tiles and separately performed instance segmentation on each tile. We explored several instance segmentation algorithms that work well on natural images and developed a set of effective methods that are applicable to remote sensing images. Then we merged the prediction results of all small tiles into seamless, continuous segmentation results through our proposed overlap-tile fusion strategy. We achieved first place among 486 teams in the challenge.

关键词： high-resolution remote sensing images instance segmentation cultivated land segmentation

Detection Transformer with Stable Matching

学校读者我要写书评

暂无评论

Detection Transformer with Stable Matching

International Conference on Computer Vision (ICCV)

作者： Shilong Liu Tianhe Ren Jiayu Chen Zhaoyang Zeng Hao Zhang Feng Li Hongyang Li Jun Huang Hang Su Jun Zhu Lei Zhang Dept. of Comp. Sci. and Tech. BNRist Center State Key Lab for Intell. Tech. & Sys. Institute for AI Tsinghua-Bosch Joint Center for ML Tsinghua University International Digital Economy Academy (IDEA) Platform of AI (PAI) Alibaba Group The Hong Kong University of Science and Technology South China University of Technology

关键词：

Bi-level score matching for learning energy-based latent variable models 34

学校读者我要写书评

暂无评论

Bi-level score matching for learning energy-based latent var...

34th Conference on Neural Information Processing systems, NeurIPS 2020

作者： Bao, Fan Li, Chongxuan Xu, Kun Su, Hang Zhu, Jun Zhang, Bo Dept. of Comp. Sci. & Tech. Institute for AI THBI Lab. BNRist Center State Key Lab for Intell. Tech. & Sys. Tsinghua University Beijing China

Score matching (SM) [24] provides a compelling approach to learn energy-based models (EBMs) by avoiding the calculation of partition function. However, it remains largely open to learn energy-based latent variable models (EBLVMs), except some special cases. This paper presents a bi-level score matching (BiSM) method to learn EBLVMs with general structures by reformulating SM as a bilevel optimization problem. The higher level introduces a variational posterior of the latent variables and optimizes a modified SM objective, and the lower level optimizes the variational posterior to fit the true posterior. To solve BiSM efficiently, we develop a stochastic optimization algorithm with gradient unrolling. Theoretically, we analyze the consistency of BiSM and the convergence of the stochastic algorithm. Empirically, we show the promise of BiSM in Gaussian restricted Boltzmann machines and highly nonstructural EBLVMs parameterized by deep convolutional neural networks. BiSM is comparable to the widely adopted contrastive divergence and SM methods when they are applicable;and can learn complex EBLVMs with intractable posteriors to generate natural images. © 2020 Neural information processing systems foundation. All rights reserved.

关键词： Convolutional neural networks

Grounding DINO: Marrying DINO with Grounded Pre-Training for Open-Set Object Detection

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Liu, Shilong Zeng, Zhaoyang Ren, Tianhe Li, Feng Zhang, Hao Yang, Jie Jiang, Qing Li, Chunyuan Yang, Jianwei Su, Hang Zhu, Jun Zhang, Lei Dept. of Comp. Sci. and Tech. BNRist Center State Key Lab for Intell. Tech. & Sys. Institute for AI Tsinghua-Bosch Joint Center for ML Tsinghua University China China The Hong Kong University of Science and Technology Hong Kong The Chinese University of Hong Kong Shenzhen China Microsoft Research Redmond United States South China University of Technology China

In this paper, we develop an open-set object detector, called Grounding DINO, by marrying Transformer-based detector DINO with grounded pre-training, which can detect arbitrary objects with human inputs such as category names or referring expressions. The key solution of open-set object detection is introducing language to a closed-set detector for open-set concept generalization. To effectively fuse language and vision modalities, we conceptually divide a closed-set detector into three phases and propose a tight fusion solution, which includes a feature enhancer, a language-guided query selection, and a cross-modality decoder for modalities fusion. We first pre-train Grounding DINO on large-scale datasets, including object detection data, grounding data, and caption data, and evaluate the model on both open-set object detection and referring object detection benchmarks. Grounding DINO performs remarkably well on all three settings, including benchmarks on COCO, LVIS, ODinW, and RefCOCO/+/g. Grounding DINO achieves a 52.5 AP on the COCO zero-shot1 detection benchmark. It sets a new record on the ODinW zero-shot benchmark with a mean 26.1 AP. We release some checkpoints and inference codes at https://***/IDEA-Research/GroundingDINO. Copyright © 2023, The Authors. All rights reserved.

关键词： Object detection

Dynamics of interfacial carriers and negative photoconductance in CH3NH3PbBr3-ZnO heterostructure

学校读者我要写书评

暂无评论

APPLIED PHYSICS LETTERS 2021年第17期118卷 171901-171901页

作者： Chen, Feng Shi, Zengliang Chen, Jinping Cui, Qiannan Jian, Aoqun Zhu, Yizhi Xu, Qingyu Lou, Zhidong Xu, Chunxiang Southeast Univ Natl Demonstrat Ctr Expt Biomed Engn Educ Sch Phys State Key Lab Bioelect Nanjing 210096 Peoples R China Nanjing Tech Univ Sch Phys & Math Sci Nanjing 211816 Peoples R China Taiyuan Univ Technol Coll Informat Engn & Comp Sci MicroNano Syst Res CtrShanxi Prov & Minist Educ Key Lab Adv Transducers & Intelligent Control Sys Taiyuan 030024 Peoples R China Beijing Jiaotong Univ Inst Optoelect Technol Key Lab Luminescence & Opt Informat Minist Educ Beijing 100044 Peoples R China

Understanding the interfacial charge transfer process and its dynamical mechanism is crucial to design efficient photoelectric devices. Methylammonium lead halide perovskite and zinc oxide (ZnO) have been demonstrated as promising candidates for excellent solar cells and photodetectors. However, the carrier transport process has not yet been fully explored on the perovskite surface and the perovskite/ZnO interface. Herein, a CH3NH3PbBr3-ZnO heterojunction was constructed as a quasi-phototransistor, where an apparent negative photoconductance was observed under illumination. Based on electrical and optical characterization, the photogenerated carrier transfer dynamics at CH3NH3PbBr3, ZnO, and their interface were investigated in detail. It can be assigned that the photogenerated electrons transfer toward the CH3NH3PbBr3 surface and the holes transfer to interior, so that the light-induced built-in electric field change would serve as the photogate to control the current flowing in the CH3NH3PbBr3-ZnO channel. These results provide clear images on the charge diffusion and drift process in the CH3NH3PbBr3-ZnO heterostructure. The study on the dynamics of negative photoconductivity of CH3NH3PbBr3-ZnO has great value for understanding the carrier transport properties and constructing perovskite heterostructure memory and optical switching devices.

关键词： Zinc oxide

Maximum Likelihood Training for Score-Based Diffusion ODEs by High-Order Denoising Score Matching

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Lu, Cheng Zheng, Kaiwen Bao, Fan Chen, Jianfei Li, Chongxuan Zhu, Jun Dept. of Comp. Sci. & Tech. Institute for AI Tsinghua-Bosch Joint Center for ML BNRist Center State Key Lab for Intell. Tech. & Sys. Tsinghua University Peng Cheng Laboratory China Gaoling School of AI Renmin University of China Beijing Key Lab of Big Data Management and Analysis Methods Beijing China

Score-based generative models have excellent performance in terms of generation quality and likelihood. They model the data distribution by matching a parameterized score network with first-order data score functions. The score network can be used to define an ODE ("score-based diffusion ODE") for exact likelihood evaluation. However, the relationship between the likelihood of the ODE and the score matching objective is unclear. In this work, we prove that matching the first-order score is not sufficient to maximize the likelihood of the ODE, by showing a gap between the maximum likelihood and score matching objectives. To fill up this gap, we show that the negative likelihood of the ODE can be bounded by controlling the first, second, and third-order score matching errors;and we further present a novel high-order denoising score matching method to enable maximum likelihood training of score-based diffusion ODEs. Our algorithm guarantees that the higher-order matching error is bounded by the training error and the lower-order errors. We empirically observe that by high-order score matching, score-based diffusion ODEs achieve better likelihood on both synthetic data and CIFAR-10, while retaining the high generation quality. Copyright © 2022, The Authors. All rights reserved.

关键词： Diffusion