Sun Yat-senuniversity

Shenzhen Sensetime Technology Co., Ltd.

Huawei Device Co., Ltd.

Liang Lin

Qing Wang

Bo Jiang

Ruimao Zhang

Chongyu Chen

Xiaogang Wang

Ping Luo

Wenxiu Sun

Shi Qiu

Yimin Chen

Yuying Ge

Baowen Zheng

Kuanze Ren

Qingxing Cao

Yukang Gan

Hongbin Sun

Xiangyu Xu

An active three-dimensional scene information acquisition method based on dimensionality-reduced sparse representation is provided. The method jointly processes multiple one-dimensional active detection signals collected synchronously to achieve three-dimensional positioning of objects in a detected scene or three-dimensional reconstruction of a scene structure. Through an active detection system equipped with one transmitter and multiple receivers, simultaneous three-dimensional positioning of multiple targets in a scene or three-dimensional reconstruction of the geometry of the scene is achieved.

Active three-dimensional scene information acquisition method based on dimensionality-reduced sparse representation

A method and apparatus for depth estimation of a monocular image, and a storage medium are provided. The method includes: obtaining, through a depth estimation neural network, a global feature of a monocular image according to absolute features of preset regions and relative features among the preset regions in the monocular image; and obtaining a predicted depth map of the monocular image according to the global feature, and the absolute features of preset regions and relative features among the preset regions in the monocular image.

Method and apparatus for depth estimation of monocular image, and storage medium

A method for recommending collocating dress includes: obtaining a first dress image from an input image; obtaining a first dress feature of the first dress image; obtaining, based on the first dress feature, a second dress feature matching the first dress feature from a collocation database, the second dress feature being a dress feature of a second dress image; and obtaining, based on the second dress feature, a dress image identical to and/or similar to the second dress image from a recommendation database as a collocating dress image. Based on the method in the embodiments, at least one collocation recommendation is provided for a dress image, and the recommendation result is more objective and is more in accordance with user demands.

Methods and apparatus for recommending collocating dress, electronic devices, and storage media

Target matching method and apparatus, electronic device, and storage medium, including: extracting feature vector of each frame in query image sequence and feature vector of each frame in candidate image sequence; determining self-expression feature vector of query image sequence, collaborative expression feature vector of the query image sequence, self-expression feature vector of candidate image sequence, and collaborative expression feature vector of candidate image sequence based on feature vector of each frame in query image sequence and feature vector of each frame in candidate image sequence; determining similarity feature vector between query image sequence and candidate image sequence based on self-expression feature vector of query image sequence, collaborative expression feature vector of query image sequence, self-expression feature vector of candidate image sequence, and collaborative expression feature vector of candidate image sequence; and determining matching result between query image sequence and candidate image sequence based on similarity feature vector.

Target matching method and apparatus, electronic device, and storage medium

Provided is an image deblurring algorithm based on a sparse positive source separation model. The image deblurring algorithm is used for performing processing on a blurry image collected by an optical microscopic imaging system and generated due to diffraction effect and optical deviation; under a condition of single light sensitive imaging and of not increasing an external imaging device, a spatial resolution of the optical microscopic system may be improved to a nanoscale. In the disclosure, a blurring process of microscopic imaging is expressed as a linear combination of a Point Spread Function of the imaging system; by embedding the process into a positive separation optimized frame, adding a sparsity constraint and solving to remove blurring, the high-resolution microscopic imaging is implemented.

Image deblurring algorithm based on sparse positive source separation model

This invention is an image retrieval method based on bit-scalable deep hashing learning. According to the method, the training images is used to generate a batch of image triples, wherein each of the triples contains two images with the same label and one image with a different label. The purpose of model training is to maximize a margin between matched image pairs and unmatched image pairs in the Hamming space. The deep convolutional neural network is utilized to train the model in an end-to-end fasion, where discriminative images features and has functions are simultaneously optimized. Furthermore, each bit of the hashing codes is unequally weighted so that we can manipulate the code length by truncating the insignificant bits. It is also shown that the generated bit-scalable hashing codes well preserve the discriminative powers with sorter code lengths.

Image retrieval method based on variable-length deep hash learning

The disclosure relates to a method for automatically recognizing liver tumor types in ultrasound images. The method specifically comprises: using a plurality of Regions of Interest (ROIs) to represent a CEUS image; different lesions are distinguished by the performance and changes of the ROI in time and space; representing a space-time relationship between ROIs by establishing a model in time and space at the same time; and determining, by the model, a relatively appropriate ROI and relevant parameters of the model according to existing CEUS lesion samples by means of an iterative learning method. After giving a sample, an appropriate ROI can be determined and a reference diagnosis for the lesion can be given by removing part of inappropriate ROIs in advance and by means of a rapid search method.

Method for automatically recognizing liver tumor types in ultrasound images

A method, a system, and a terminal for performing a system update between mobile communication terminals. The method includes: determining, by a target mobile terminal, whether hardware is compatible with a source mobile terminal; if the hardware is compatible between both terminals, obtaining, by the target mobile terminal through a wireless network, a system backup file transferred by the source mobile terminal; and performing, by the target mobile terminal, a system update according to the system backup file. A system update can be implemented between mobile phones through a wireless network without depending on an update server or downloading an update file from the update server, thereby reducing the burden of the update server, and facilitating exchange of systems between mobile phone users.

Growing community of inventors

Shenzhen, China

Liang Lin