|
|
|
| Finding Task-Relevant Features for Few-Shot Learning by Category Traversal |
中用类别遍历法寻找任务相关特征 |
|
| Edge-Labeling Graph Neural Network for Few-Shot Learning |
用于少镜头学习的图神经网络 |
|
| Generating Classification Weights With GNN Denoising Autoencoders for Few-Shot Learning |
用生成实现少镜头学习 |
|
| Kervolutional Neural Networks |
神经网络 |
传统卷积运算在神经网络中的扩展——（Kernel Convolution）：对于传统卷积的非线性化——使用非线性映射(根据输入特性和卷积核)卷积(公式4) |
| Why ReLU Networks Yield High-Confidence Predictions Far Away From the Training Data and How to Mitigate the Problem |
为什么产生的以及如何缓解问题 |
|
| On the Structural Sensitivity of Deep Convolutional Networks to the Directions of Fourier Basis Functions |
深度卷积网络对傅立叶基函数方向的 |
|
| Neural Rejuvenation: Improving Deep Network Training by Enhancing Computational Resource Utilization |
神经再生：通过提高计算资源利用率改进深度网络训练 |
|
| Hardness-Aware Deep Metric Learning |
深度测量学习 |
|
| Auto-DeepLab: Hierarchical Neural Architecture Search for Semantic Image Segmentation |
Auto-DeepLab：语义图像分割的层次神经结构搜索 |
神经网络的自动搜索优化（而不是预先定义） |
| Learning Loss for Active Learning |
的学习损失 |
|
| Striking the Right Balance With Uncertainty |
以不确定性达到正确的平衡 |
|
| AutoAugment: Learning Augmentation Strategies From Data |
自增强：从数据中学习增强策略 |
|
| SDRSAC: Semidefinite-Based Randomized Approach for Robust Point Cloud Registration Without Correspondences |
SDRSAC：基于半定的随机方法实现点云配准 |
|
| BAD SLAM: Bundle Adjusted Direct RGB-D SLAM |
BAD SLAM：Bundle Adjusted直接RGB-D SLAM |
提出了的BA方法（传统的密集BA方法比较耗时） 算法贡献主要在于提出使用的概念，从而利用Surfel来估计一组像素，因而达到密集BA的目的 代价函数见公式1，BA优化算法见Algo.1 代码： |
| Revealing Scenes by Inverting Structure From Motion Reconstructions |
通过structure From Motion重建反转来显示场景 |
|
| Strand-Accurate Multi-View Hair Capture |
精确的头发捕捉 |
|
| DeepSDF: Learning Continuous Signed Distance Functions for Shape Representation |
deepSDF：学习连续符号距离函数的形状表示 |
|
| Pushing the Boundaries of View Extrapolation With Multiplane Images |
使用多平面图像视图外推边界 |
|
| GA-Net: Guided Aggregation Net for End-To-End Stereo Matching |
GA-Net：端到端立体匹配的引导聚合网 |
提出两种方法：semi-global和local，分别对应无纹理区和细结构/边缘区 |
| Real-Time Self-Adaptive Deep Stereo |
实时自适应深度立体 |
：来解决问题（训练集为合成数据，而真实测试集为真实场景）。在实际使用中，每帧数据（对）不仅用来计算视差，，达到自适应的目的 |
| LAF-Net: Locally Adaptive Fusion Networks for Stereo Confidence Estimation |
LAF-Net：用于立体置信估计的局部(L)自适应(A)融合(F)网络 |
（Confidence map）用以衡量每个点的（估计后）视差的置信度（如图1），进而对不同置信度像素点的视差可以refine等后处理。 |
| NM-Net: Mining Reliable Neighbors for Robust Feature Correspondences |
NM-Net：挖掘可靠的邻域，以实现强大的特征对应 |
特征点对应一般有SIFT等局部特征对应，但是初始化的对应特征点不可避免包含错误的对应，因此需要来“”正确的对应特征点。本文主要关注基于学习的方法，来实现正确地“选择”对应特征点。 |
| Coordinate-Free Carlsson-Weinshall Duality and Relative Multi-View Geometry |
Carlsson-Weinshall对偶及相对多视图几何 |
|
| Deep Reinforcement Learning of Volume-Guided Progressive View Inpainting for 3D Point Scene Completion From a Single Depth Image |
利用深度强化学习实现的基于体引导渐进视图修补的三维点场景补全 |
|
| Video Action Transformer Network |
视频动作转换网络 |
|
| Timeception for Complex Action Recognition |
复杂动作识别的时间感知 |
|
| STEP: Spatio-Temporal Progressive Learning for Video Action Detection |
STEP：视频动作检测的时空渐进学习 |
|
| Relational Action Forecasting |
预测 |
|
| Long-Term Feature Banks for Detailed Video Understanding |
详细视频理解的长期功能库 |
|
| Which Way Are You Going? Imitative Decision Learning for Path Forecasting in Dynamic Scenes |
你往哪边走？路径预测的模拟决策学习 |
|
| What and How Well You Performed? A Multitask Learning Approach to Action Quality Assessment |
你的表现如何？行动质量评估的多任务学习方法 |
|
| MHP-VOS: Multiple Hypotheses Propagation for Video Object Segmentation |
MHP-VOS：视频对象分割的多假设传播 |
|
| 2.5D Visual Sound |
2.5D视觉声音 |
|
| Language-Driven Temporal Activity Localization: A Semantic Matching Reinforcement Learning Model |
语言驱动的时间活动定位：语义匹配的强化学习模型 |
|
| Gaussian Temporal Awareness Networks for Action Localization |
用于动作定位的高斯时间感知网络 |
|
| Efficient Video Classification Using Fewer Frames |
使用更少帧的视频分类 |
|
| Parsing R-CNN for Instance-Level Human Analysis |
解析R-CNN实现实例级的人分析 |
|
| Large Scale Incremental Learning |
增量学习 |
增量学习：不断增加新类别的学习。由于不断增加新类别，导致旧类别的样本减少，造成数据不平衡，从而使得旧类别的识别度下降。本文关注类别不平衡问题的解决 |
| TopNet: Structural Point Cloud Decoder |
TopNet：点云解码器 |
|
| Perceive Where to Focus: Learning Visibility-Aware Part-Level Features for Partial Person Re-Identification |
感知关注点：学习可见性感知特征实现部分人重识别 |
|
| Meta-Transfer Learning for Few-Shot Learning |
元转移学习实现少镜头学习 |
|
| Structured Binary Neural Networks for Accurate Image Classification and Semantic Segmentation |
用于精确图像分类和语义分割的结构化二元神经网络 |
由原始网络经过网络结构改进及权值二元化，实现 |
| Deep RNN Framework for Visual Sequential Applications |
用于视觉序列应用的深度RNN框架 |
|
| Graph-Based Global Reasoning Networks |
全局 |
通过引入全局信息，。如图1,2，首先将空间（笛卡尔坐标）像素投影到交互空间（interaction space），在交互空间通过全连接（图）网络，获取全局信息，然后再反投影到原始空间。 |
| SSN: Learning Sparse Switchable Normalization via SparsestMax |
SSN：通过SparsestMax学习稀疏可切换 |
|
| Spherical Fractal Convolutional Neural Networks for Point Cloud Recognition |
用于点云识别的球形分形 |
|
| Learning to Generate Synthetic Data via Compositing |
学习通过合成生成 |
|
| Divide and Conquer the Embedding Space for Metric Learning |
划分并征服实现度量学习 |
|
| Latent Space Autoregression for Novelty Detection |
新颖性检测的潜在空间自回归 |
|
| Attending to Discriminative Certainty for Domain Adaptation |
注意判别确定性实现域适应 |
|
| Feature Denoising for Improving Adversarial Robustness |
特征去噪提高 |
|
| Selective Kernel Networks |
核网络 |
|
| On Implicit Filter Level Sparsity in Convolutional Neural Networks |
隐式滤波级 |
研究比较网络中采用不同方法（正则、优化等）情形下的网络系数稀疏性情况 |
| FlowNet3D: Learning Scene Flow in 3D Point Clouds |
FlowNet3D：学习场景流 |
|
| Scene Memory Transformer for Embodied Agents in Long-Horizon Tasks |
基于场景记忆变换器的嵌入式代理 |
|
| Co-Occurrent Features in Semantic Segmentation |
语义分割中的共现特征 |
考虑的语义分割中不同语义之间的关系（共现：，图3），实际上是考虑不同位置之间的点积信息 |
| Bag of Tricks for Image Classification with Convolutional Neural Networks |
图像分类中采用的技巧 |
|
| Learning Channel-Wise Interactions for Binary Convolutional Neural Networks |
二元卷积神经网络的通道交互学习 |
|
| Knowledge Adaptation for Efficient Semantic Segmentation |
有效语义分割的知识自适应 |
基于的方法（利用复杂的teacherNet指导简单的studentNet，从而得到更快速、效果更佳的推断），实现语义分割 |
| Parametric Noise Injection: Trainable Randomness to Improve Deep Neural Network Robustness Against Adversarial Attack |
参数噪声注入：可训练的以深度神经网络对抗攻击的鲁棒性 |
|
| Invariance Matters: Exemplar Memory for Domain Adaptive Person Re-Identification |
不变性问题：基于范例记忆的域适应人再识别 |
同时利用source域带标签的训练样本和target域无标签的训练样本，训练具备的跨域ReID。如图2，其中target域样本考虑，形成，辅助训练 |
| Dissecting Person Re-Identification From the Viewpoint of Viewpoint |
从视角的视角剖析人再识别 |
两点贡献：1.提出了一个构建不同视角训练集的引擎（算法）；2. 分析不同视角对ReID的影响 |
| Learning to Reduce Dual-Level Discrepancy for Infrared-Visible Person Re-Identification |
学习减少双级差异实现人再识别 |
红外图像的ReID，两两差异（discrepancy），采用两个不同子网来处理 |
| Progressive Feature Alignment for Unsupervised Domain Adaptation |
基于渐进特征对齐的域自适应 |
|
| Feature-Level Frankenstein: Eliminating Variations for Discriminative Recognition |
特征级Frankenstein：基于差异消除的判别性识别 |
|
| Learning a Deep ConvNet for Multi-Label Classification With Partial Labels |
基于深度ConvNet学习的多标签分类 |
|
| Generalized Intersection Over Union: A Metric and a Loss for Bounding Box Regression |
联合上的广义交集：用于BoundingBox回归的度量和损失 |
|
| Densely Semantically Aligned Person Re-Identification |
基于密集语义对齐的人再识别 |
首先利用DensePose模型，将人体进行语义分割（24种语义），然后对于分割后的人体部分进行对齐（alignmeng）。最后，将这些24幅对齐后的图像组作为输入，输入到辅助网络中，帮助提高主网的ReID能力（图3） |
| Generalising Fine-Grained Sketch-Based Image Retrieval |
基于细粒度草图的图像检索 |
|
| Adapting Object Detectors via Selective Cross-Domain Alignment |
选择性跨域对齐实现目标检测器调整 |
|
| Cyclic Guidance for Weakly Supervised Joint Detection and Segmentation |
基于循环引导的联合检测与分割 |
|
| Thinking Outside the Pool: Active Training Image Creation for Relative Attributes |
池化外思维：基于主动训练图像创建的相关属性 |
|
| Generalizable Person Re-Identification by Domain-Invariant Mapping Network |
基于域不变映射网络的人再识别 |
利用多个domain的数据训练，得到domain可推广的ReID（新的domain无需再update）。采用元学习的思想，网络图见图1 |
| Visual Attention Consistency Under Image Transforms for Multi-Label Image Classification |
视觉注意一致性实现多标签图像分类 |
|
| Re-Ranking via Metric Fusion for Object Retrieval and Person Re-Identification |
基于的实现目标检索和人再识别 |
人再识别后的，将几种Fusion算法统一起来。 目标函数：公式10 |
| Unsupervised Open Domain Recognition by Semantic Discrepancy Minimization |
基于语义差异最小化的开放域识别 |
|
| Weakly Supervised Person Re-Identification |
人再识别 |
所谓“弱监督”，在这里指的是Gallery集合中的是视频帧，每帧有若干个人，而对于标签只指出含有哪些人，而不指出对应哪个人。Probe只单个人的patch，且标签为确定的人。这是一个的问题 |
| PointRCNN: 3D Object Proposal Generation and Detection From Point Cloud |
PointRCNN：从实现三维对象Proposal和 |
|
| Automatic Adaptation of Object Detectors to New Domains Using Self-Training |
利用自训练使自动 |
|
| Deep Sketch-Shape Hashing With Segmented 3D Stochastic Viewing |
基于分段三维随机视图的深度 |
|
| Generative Dual Adversarial Network for Generalized Zero-Shot Learning |
基于生成对偶对抗网络的广义 |
|
| Query-Guided End-To-End Person Search |
基于查询引导的人员搜索 |
|
| Libra R-CNN: Towards Balanced Learning for Object Detection |
Libra R-CNN：的平衡学习 |
在RNN网络中，作者认为存在（采样不平衡、各层级特征不平衡、损失函数中各项之间不平衡），从而导致效果下降。本文在网络中的不同位置，添加不同的（） 效果有两个点的提升（表1），代码：https://github.com/OceanPang/Libra_R-CNN |
| Learning a Unified Classifier Incrementally via Rebalancing |
通过重新平衡实现的逐步学习 |
|
| Feature Selective Anchor-Free Module for Single-Shot Object Detection |
基于特征选择无锚模块的目标检测 |
|
| Bottom-Up Object Detection by Grouping Extreme and Center Points |
通过对极值点和中心点进行分组的目标检测 |
|
| Feature Distillation: DNN-Oriented JPEG Compression Against Adversarial Examples |
特征蒸馏：基于DNN的JPEG压缩与对抗性示例 |
|
| SCOPS: Self-Supervised Co-Part Segmentation |
SCOPS：共部分分割 |
|
| Unsupervised Moving Object Detection via Contextual Information Separation |
基于上下文信息分离的无监督运动目标检测 |
|
| Pose2Seg: Detection Free Human Instance Segmentation |
Pose2Seg： |
专门针对下的人，利用预测 |
| DrivingStereo: A Large-Scale Dataset for Stereo Matching in Autonomous Driving Scenarios |
驾驶立体：用于的大规模数据集 |
|
| PartNet: A Large-Scale Benchmark for Fine-Grained and Hierarchical Part-Level 3D Object Understanding |
PartNet：一个用于Part-Level三维对象理解的大规模基准 |
|
| A Dataset and Benchmark for Large-Scale Multi-Modal Face Anti-Spoofing |
大型多模人脸防欺骗的数据集与基准 |
|
| Unsupervised Learning of Consensus Maximization for 3D Vision Problems |
共识最大化的无监督学习 |
|
| VizWiz-Priv: A Dataset for Recognizing the Presence and Purpose of Private Visual Information in Images Taken by Blind People |
VizWiz-Priv：一个数据集，用于识别盲人拍摄的图像中私人视觉信息的存在和目的。 |
|
| Structural Relational Reasoning of Point Clouds |
结构关系推理 |
|
| MVF-Net: Multi-View 3D Face Morphable Model Regression |
MVF-Net：人脸形态模型回归 |
|
| Photometric Mesh Optimization for Video-Aligned 3D Object Reconstruction |
光度网格优化实现三维对象重建 |
|
| Guided Stereo Matching |
引导 |
（可以容易地转化为对应点的视差值），利用这部分信息作为引导，辅助实现立体视觉（公式1-4，通过图2b,c可以看出，其对性能的提升也是有明显的好处的） |
| Unsupervised Event-Based Learning of Optical Flow, Depth, and Egomotion |
光流、深度和自我学习 |
|
| Modeling Local Geometric Structure of 3D Point Clouds Using Geo-CNN |
基于Geo-CNN的局部几何结构建模 |
|
| 3D Point Capsule Networks |
的胶囊网络 |
|
| GS3D: An Efficient 3D Object Detection Framework for Autonomous Driving |
GS3D：一种高效的自动驾驶三维框架 |
|
| Single-Image Piece-Wise Planar 3D Reconstruction via Associative Embedding |
基于关联嵌入的平面三维重建 |
|
| 3DN: 3D Deformation Network |
3DN:3D变形网络 |
|
| HorizonNet: Learning Room Layout With 1D Representation and Pano Stretch Data Augmentation |
HorizonNet：基于一维表示和Pano拉伸数据扩充的室布局学习 |
|
| Deep Fitting Degree Scoring Network for Monocular 3D Object Detection |
基于深度拟合度评分网络的三维目标检测 |
|
| Pushing the Envelope for RGB-Based Dense 3D Hand Pose Estimation via Neural Rendering |
利用神经渲染实现手部姿态估计 |
|
| Self-Supervised Learning of 3D Human Pose Using Multi-View Geometry |
基于多视图几何的自监督学习 |
|
| FSA-Net: Learning Fine-Grained Structure Aggregation for Head Pose Estimation From a Single Image |
FSA-Net：细粒度学习实现头部姿势估计 |
|
| Dense 3D Face Decoding Over 2500FPS: Joint Texture & Shape Convolutional Mesh Decoders |
2500fps以上密集三维人脸解码：联合纹理和形状解码器 |
|
| Does Learning Specific Features for Related Parts Help Human Pose Estimation? |
学习特定特征是否有助于
标签： fa传感器c2dm
|