计算机视觉应用技术
2020-05-17 17:46:50 1 举报AI智能生成
计算机视觉应用领域与方法
作者其他创作
大纲/内容
深度估计
method
多目深度估计
GeoNet(Geometric Neural Network)
单目深度估计
Deep Ordinal Regression Network(DORN)
metric
Roor Mean Squared Error(RMSE)
Lg Error(LG)
图像检索
method
SPoC(sum-pooled convolutional features)
Vector of Local Aggregated Descriptor（VLAD）
DELF(Attentive Deep Local Features)
Deep Local and Global Features(DLGF)
metric
MAP(mean average precision)
Rank-k
CMC(culmulative matching curve)
图像描述
metric
BLEU(Bilingual Evaluation Understudy)
CIDER(Consensus-based Image Description Evaluation)
METEOR(Metric for Evaluation of Translation with Explicit ORdering)
ROUGE(Recall-Oriented Understudy for Gisting Evaluation)
SPICE(Semantic Propositional Image Caption Evaluation)
Perplexity
method
m-RNN(Multimodal Recurrent Neural Networks)
NIC(Neural Image Caption Generator)
Adaptive Attention Model
Show and Tell
dataset
The Visual Genome Dataset
ABSTRACT-50S dataset
子主题
视觉问答
dataset
DAQUAR (DAtaset for QUestion Answering on Real-world images)
MS-COCO-QA(Question Answering)
metric
Wu-Palmer similarity(WUPS)
Resnik Similarity
Lin Similarity
method
Neural-Image-QA
Movie Question Answering (MQA)
图像修复
Context Encoders: Feature Learning by Inpainting
DCGAN(Deep Convolution)
图像配准
method
STN(Spatial Transofrom Network) - TPN(Thin Plate Network)
DIRNet(Deformable Image Resistration)
SGM(Semi Global Matching)
Graph Based Matching Method
metric
repeatability
文字识别
method
文本检测
CTPN(Connectionist Text Proposal Network)
PSENet(Progressive Scale Expansion Network)
EAST(Efficient and Accurate Scene Text Detector)
Single-Shot Arbitrarily-Shaped Text Detector(SAST)
Differentiable Binarization（DB）
方向检测
anglenet
文本识别
CRNN(Convolutional Recurrent Neural Network)
Rosetta
SpaTial Attention Residue Network（STAR-Net)
Robust Scene Text Recognition with Automatic Rectification(RARE)
Semantic Reasoning Networks（SRN）
metric
图像去噪
Bayesian Least Squares Gaussian scale mixture (BLS-GSM)
Block-matching and 3D filtering (BM3D)
cascade of shrinkage fields (CSF)
trainable nonlinear reaction diffusion (TNRD)
线条检测
method
VPGNet(Vanishing Point Guided Network)
SCNN(Spatial CNN for Traffic Scene Understanding)
PINet(Point Intance Network)
Hough Transoform
metric
mIOU
3D视觉点云
人脸关键点
method
Convolutional Pose Machines（CPM）
RMPE( Regional Multi-person Pose Estimation)
Practical Facial Landmark Detector（PFLD）
metric
OKS（object keypoints）
dataset
LSP（Leeds Sports Pose Dataset）
FLIC（Frames Labeled In Cinema）
MPII（MPII Human Pose Dataset）
神经渲染
Neural 3D Mesh Renderer:NMR
DIB-R: Interpolation-based Differentiable Renderer
3D视觉影像
single image for kenburns effect
开拓思路
强化学习
图神经网络
多模态
超分重建
method
视频超分辨率重建
FRVSR (Frame-Recurrent Video Super-Resolution)
EDCN (Enhanced Deformable Convolutional Networks)
RBPN (Recurrent Back-Projection Network)
图像超分辨率重建
ESRGAN(Enhanced SRGAN)
TecoGAN (TEmporally COherent GAN)
ESPCN (Efficient Sub-Pixel Convolutional Neural Network)
DNI (Deep network interpolation)
RCAN (Residual Channel Attention Networks)
SAN (Second-Order Attention Network)
metric
Meansquareerror（MSE)
Signalnoiseratio（SNR)
Structuresimilarity（SSIM)
样本困难
meta learning
Optimization-Based
Metric-Based
Model-Based
zero-shot learning
Siamese Neural Networks
few-shot learning
self-supervised learning
method
AMDIM/DIM (Augmented Multiscale Deep InfoMax)
CPC (Contrastive Predictive Coding)
CMC (Contrastive Multi-view Coding)
SimCLR (Simple Contrastive Learning)
PIRL (Pretext-Invariant Representations Learning)
MoCo (Momentum Contrast)
unsupervised learning
unsupervised learning is similar to self-supervised learning except in labeling way or no label at all
目标检测
metric
IOU
mAP
PR(Precision Recall Curve)
method
双阶段目标检测
Faster RCNN(Region CNN)
RetinaNet
Scale Normalization for Image Pyramids (SNIP)
Region-based Fully Convolutional Networks(RFCN)
单阶段目标检测
SSD(Single Shot Detection)
YOLO(Yon Onluy Look Once) series
多阶段目标检测
Cascade RCNN
Anchor Free Detection
语义分割
metirc
pixel accuracy
Mean IOU
method
FCN (Fully Convolution Network)
UNet ( Medical Image Segmentation)
CCNet( Criss-Cross Attention)
low-rank-recovery network (LRRNet)
DANet（Dual Attention Network）
SegNet ( semantic pixel-wise segmentation)
DeepLab Series
实例分割
method
单阶段实例分割
FCIS (Fully Convolutional Instance-aware Semantic Segmentation)
YOLACT(You Only Look At CoefficienTs)
PolarMask
BlendMask
SOLO (Segmenting Objects by Locations)
RDSNet (Reciprocal Object Detection and Instance Segmentation.)
双阶段实例分割
Mask-RCNN
metric
Pixel Accuracy(PA）
Mean Pixel Accuracy(MPA)
目标追踪
metric
Expected average overlap (EAO)
method
Oblique Random forest (Obli-Raf)
ensemble-based tracking (EBT)
Multiple Experts Using Entropy (MEEM)
consistent low-rank sparse tracker(CLRST)
Continuous Convolution Operator Tracker (C-COT)
Efficient Convolution Operators（ECO）
MCCF（Multi-Channel Correlation Filters）
(SANet) Structure-Aware Network for Visual Tracking
(VITAL) VIsual Tracking via Adversarial Learning
(GOTURN) Generic Object Tracking Using Regression Networks
(CREST) Convolutional Residual Learning for Visual Tracking
ATOM (Accurate Tracking by Overlap Maximization)
Tips
online learning(cls) & offline learning(bbox)
图像生成
method
CRN(Convolutional Recurrent Network)
SPADE(Semantic Image Synthesis with Spatially-Adaptive Normalization.)
Pix2Pix
metric
IS(Inception Score)
FID(Frechet Inception Distance)
度量学习
method
Large margin nearest neighbor (LMNN)
Logistic Discriminant Metric Learning (LDML)
Information Theoretic Metric Learning (ITML)
proxy-nca(s Neighborhood Component Analysis)
metric
pair ranking loss (Siamese Net)
triplet ranking loss (triplet mining)
dataset
SOP(standford online product)
CUB200(Caltech-UCSD Birds 200 )
Cars(196)
全景分割
metric
PQ（Panoptic Quality）
RQ（recognition quality）
SQ（segmentation quality）
method
TASCNet(Things and Stuff Consistency)
姿态估计
method
densepose
alphapose
openpose
deeppose
dataset
MPII(Multi Person)
metric
OKS(object keypoint similarity)
PCK(Percentage of Correct Keypoints)
图像分类
method
多标签图像分类
HCP(cross-hypothesis max-pooling)
Regional Latent Semantic Dependencies (RLSD)
单标签图像分类
BlockQNN(Block-wise Neural Network Architecture Generation Q-learning)
CRU-Net (Collective Residual Networks)
IGCV(Interleaved Low-Rank Group Convolutions)
细粒度图像分类
metric
Recall
Preicesion
F-Score
人脸识别
method
检测
对齐
验证
识别
metric
False Accept Rate(FAR)
False Reject Rate (FRR)
0 条评论
下一页
为你推荐
查看更多
