Processing

Please wait...

Settings

Settings

Goto Application

1. CN110209859 - Place identification and model training method and device and electronic equipment

Office
China
Application Number 201910390693.2
Application Date 10.05.2019
Publication Number 110209859
Publication Date 06.09.2019
Publication Kind A
IPC
G06F 16/58
GPHYSICS
06COMPUTING; CALCULATING OR COUNTING
FELECTRIC DIGITAL DATA PROCESSING
16Information retrieval; Database structures therefor; File system structures therefor
50of still image data
58Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
G06F 16/587
GPHYSICS
06COMPUTING; CALCULATING OR COUNTING
FELECTRIC DIGITAL DATA PROCESSING
16Information retrieval; Database structures therefor; File system structures therefor
50of still image data
58Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
587using geographical or spatial information, e.g. location
G06F 16/583
GPHYSICS
06COMPUTING; CALCULATING OR COUNTING
FELECTRIC DIGITAL DATA PROCESSING
16Information retrieval; Database structures therefor; File system structures therefor
50of still image data
58Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
583using metadata automatically derived from the content
G06N 3/04
GPHYSICS
06COMPUTING; CALCULATING OR COUNTING
NCOMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS
3Computer systems based on biological models
02using neural network models
04Architecture, e.g. interconnection topology
CPC
G06F 16/5866
GPHYSICS
06COMPUTING; CALCULATING; COUNTING
FELECTRIC DIGITAL DATA PROCESSING
16Information retrieval; Database structures therefor; File system structures therefor
50of still image data
58Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
5866using information manually generated, e.g. tags, keywords, comments, manually generated location and time information
G06F 16/587
GPHYSICS
06COMPUTING; CALCULATING; COUNTING
FELECTRIC DIGITAL DATA PROCESSING
16Information retrieval; Database structures therefor; File system structures therefor
50of still image data
58Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
587using geographical or spatial information, e.g. location
G06F 16/583
GPHYSICS
06COMPUTING; CALCULATING; COUNTING
FELECTRIC DIGITAL DATA PROCESSING
16Information retrieval; Database structures therefor; File system structures therefor
50of still image data
58Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
583using metadata automatically derived from the content
G06N 3/0454
GPHYSICS
06COMPUTING; CALCULATING; COUNTING
NCOMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS
3Computer systems based on biological models
02using neural network models
04Architectures, e.g. interconnection topology
0454using a combination of multiple neural nets
G06K 9/6243
GPHYSICS
06COMPUTING; CALCULATING; COUNTING
KRECOGNITION OF DATA; PRESENTATION OF DATA; RECORD CARRIERS; HANDLING RECORD CARRIERS
9Methods or arrangements for reading or recognising printed or written characters or for recognising patterns, e.g. fingerprints
62Methods or arrangements for recognition using electronic means
6217Design or setup of recognition systems and techniques; Extraction of features in feature space; Clustering techniques; Blind source separation
6232Extracting features by transforming the feature space, e.g. multidimensional scaling; Mappings, e.g. subspace methods
624based on a separation criterion, e.g. independent component analysis
6243of decorrelation or non-stationarity, e.g. minimising lagged cross-correlations
G06K 9/6232
GPHYSICS
06COMPUTING; CALCULATING; COUNTING
KRECOGNITION OF DATA; PRESENTATION OF DATA; RECORD CARRIERS; HANDLING RECORD CARRIERS
9Methods or arrangements for reading or recognising printed or written characters or for recognising patterns, e.g. fingerprints
62Methods or arrangements for recognition using electronic means
6217Design or setup of recognition systems and techniques; Extraction of features in feature space; Clustering techniques; Blind source separation
6232Extracting features by transforming the feature space, e.g. multidimensional scaling; Mappings, e.g. subspace methods
Applicants TENCENT TECHNOLOGY (SHENZHEN) CO., LTD.
腾讯科技(深圳)有限公司
Inventors BAI DONGDONG
白栋栋
LING YONGGEN
凌永根
LIU WEI
刘威
Agents 深圳市隆天联鼎知识产权代理有限公司 44232
Title
(EN) Place identification and model training method and device and electronic equipment
(ZH) 地点识别及其模型训练的方法和装置以及电子设备
Abstract
(EN) The invention discloses a place identification and model training method and device, a computer readable storage medium and electronic equipment. The place identification and model training method comprises the steps: extracting local features of a sample image based on a first part of a CNN model; aggregating the local features into feature vectors having a first dimension based on a second portion of the CNN model; obtaining a compressed representation vector of the feature vector based on a third portion of the CNN model, the compressed representation vector having a second dimension that is less than the first dimension; and adjusting the model parameters of the first to third parts with the target of minimizing the distance between the compressed representation vectors corresponding to the plurality of images until a CNN model meeting a preset condition is obtained. According to the model training method provided by the embodiment of the invention, the compression process with trainable parameters is introduced into the CNN model, and the end-to-end training place identification model can be truly realized, and the obtained CNN model can directly obtain low-dimensional image features, so that the place identification performance is improved.
(ZH) 本发明揭示了一种地点识别及其模型训练的方法和装置、计算机可读存储介质以及电子设备。方法包括:基于CNN模型的第一部分提取样本图像的局部特征;基于CNN模型的第二部分将局部特征聚合成具有第一维数的特征向量;基于CNN模型的第三部分得到特征向量的压缩表示向量,压缩表示向量具有小于第一维数的第二维数;以及以使得多个图像对应的压缩表示向量之间的距离最小化为目标,调整第一至第三部分的模型参数,直至得到满足预设条件的CNN模型。本发明实施例提供的模型训练方法,通过在CNN模型中引入参数可训练的压缩过程,能够真正实现端到端的训练地点识别模型,得到的CNN模型能够直接获得低维度的图像特征,从而提高地点识别的性能。
Related patent documents