Wuyi University

Yikui Zhai

Junying Gan

Wenbo Deng

Junying Zeng

Qirui Ke

Ying Xu

Tianlei Wang

Zilu Ying

Chuanbo Qin

He Cao

Guohui He

Yihang Zhi

Chaoyun Mai

Cuilin Yu

Wenlue Zhou

Yu Zheng

Zhenfeng Bai

Jiangyuan Chen

Yanyang Liang

Jingling Zhang

Yueting Wu

Na Xiao

Jihua Zhou

Rong Fu

Fan Wang

Xi Wu

Zijian Li

Zhongxin Yu

Junjie Xu

Li Xiang

Bicheng Wu

Zhiyong Hong

Simian Chen

Boyuan Zhu

Yiren Long

Junda Deng

Jiongzhi Qiu

Chaojian Xie

Jingming Zhu

Jinxin Wang

Wenlve Zhou

Yijia Deng

Huixin Guan

A ball picking robot and a ball unloading method using the same. The ball picking robot includes a chassis, an L-shaped hanging lug, a ball storage frame, an electric push rod disposed on the chassis, a baffle disposed on the chassis, a bracket disposed on the chassis and a drive motor for driving the electric push rod to move. A movable end of the electric push rod is connected with one end of the L-shaped hanging lug, the other end of the L-shaped hanging lug is connected with a side of the ball storage frame, and a camera is disposed on the bracket.

Ball picking robot and ball unloading method using the same

A multi-dimensional task facial beauty prediction method and system, and a storage medium are disclosed. The method includes the steps of: at a training phase, using first facial images to optimize a shared feature extraction network for extracting shared features and to train a plurality of sub-task networks for performing facial beauty classification tasks; at a testing phase, extracting shared features of second facial images; inputting the shared features to the trained plurality of sub-task networks; and obtaining a first beauty prediction result based on first output results of the plurality of sub-task networks.

Multi-dimensional task facial beauty prediction method and system, and storage medium

Disclosed are a method and device for predicting face beauty based on migration and weak supervision and a storage medium. The method includes: preprocessing an inputted face image; training a source domain network by using the preprocessed image, and migrating a parameter of the source domain network to a target domain network; inputting a noise image marked with a noise label and a truth-value image marked with a truth-value label into the target domain network to obtain an image feature; and inputting the image feature into a classification network to obtain a final face beauty prediction result.

Method and device for predicting beauty based on migration and weak supervision, and storage medium

A beauty prediction method and device based on multitasking and weak supervision, and a storage medium are disclosed. The method includes the steps of pre-processing inputted facial images; allocating the pre-processed images to multiple tasks; extracting shared image features; and obtaining a plurality of classification results via a plurality of classification networks each including a residual network, a standard neural network and a classifier.

Beauty prediction method and device based on multitasking and weak supervision, and storage medium

Disclosed are a facial beauty prediction method and device based on multi-task migration. The method includes: performing similarity measurement based on a graph structure on a plurality of tasks to obtain an optimal combination of the plurality of tasks; constructing a facial beauty prediction model including a feature sharing layer based on the optimal combination; migrating feature parameters of an existing large-scale facial image network to the feature sharing layer of the facial beauty prediction model; inputting facial images for training to pre-train the facial beauty prediction model; and inputting a facial image to be tested to the trained facial beauty prediction model to obtain facial recognition results.

Facial beauty prediction method and device based on multi-task migration

A method for measuring an antenna downtilt angle based on a multi-scale deep semantic segmentation network is disclosed, including: collecting base station antenna data by using an unmanned aerial vehicle, and labeling an acquired antenna image with a labeling tool to make a data set; calling the data set for training and debugging a model; recognizing and detecting a target antenna, performing semantic segmentation on an output image, finally obtaining a target image finally segmented, and calculating a downtilt angle of the target image. The method is highly applicable, cost-effective, and safe.

Method for measuring antenna downtilt angle based on multi-scale deep semantic segmentation network

Disclosed are a method and apparatus for SAR image data enhancement, and a storage medium. The method includes: processing an SAR target image by electromagnetic simulation to acquire an SAR electromagnetic simulation image; and processing the SAR electromagnetic simulation image and the SAR target image by a generative adversarial network to obtain a set of virtual samples of the SAR target image.

Method and apparatus for SAR image data enhancement, and storage medium

Disclosed are a method and an apparatus for end-to-end SAR image recognition, and a storage medium. According to the disclosure, a generative adversarial network is used to enhance data and improve data richness of a SAR image, which is beneficial to subsequent network training; a semantic feature enhancement technology is also introduced to enhance semantic information of a SAR deep feature by a coding-decoding structure, which improves performances of SAR target recognition; and meanwhile, an end-to-end SAR image target recognition model with high integrity for big scenes like the Bay Area is constructed, which is helpful to improve a synthetic aperture radar target recognition model for big scenes like the Bay Area from local optimum to global optimum, increases the stability and generalization ability of the model, reduces the network complexity, and improves the target recognition accuracy.

Method and apparatus for end-to-end SAR image recognition, and storage medium

The present disclosure relates to a method for enhancing sematic features of SAR image oriented small set of samples, comprising: acquiring a sample set of an SAR target image, and performing transfer learning and training on the sample set to obtain a initialized deep neural network of an SAR target image, the sample set comprising an SAR target image and an SAR target virtual image; performing network optimization on the deep neural network by an activation function, and extracting features of the SAR target image by the optimized deep neural network to obtain a feature map; and mapping, by an auto-encoder, the feature map between a feature space and a semantic space to obtain a deep visual feature with an enhanced semantic feature.

Method and apparatus for enhancing semantic features of SAR image oriented small set of samples

Disclosed is a method for vein recognition, the method includes: performing a difference operation and a channel connection on two to-be-verified target vein images respectively to obtain a difference image and a two-channel image of the two target vein images; performing the channel connection on the obtained difference image and two-channel image to obtain a three-channel image, so as to use the three-channel image as an input of a CNN network; fine-tuning a pre-trained model SqueezeNet that completes training on an ImageNet; integrating the difference image and the three-channel image through a cascade optimization framework to obtain a recognition result; regarding a pair of to-be-verified images as a sample, transforming the sample, taking the transformed sample as the input of the CNN network, obtaining a recognition result by supervised training on the network.

Method for vein recognition, and apparatus, device and storage medium thereof

The present disclosure discloses an error modeling method and device for prediction context of reversible image watermarking. A predictor based on omnidirectional context is established; then, the prediction context is self-adaptively error modeled to obtain a self-adaptive error model; and finally, output data from the self-adaptive error model is fed back to the predictor to update and correct the prediction context, so as to correct a prediction value of a current pixel x[i,j]. Since the non-linear correlation between the current pixel and the prediction context thereof, i.e., the non-linear correlation redundancy between pixels can be found by the error modeling of the prediction context of the predictor, the non-linear correlation redundancy between the pixels can be effectively removed. Thus, the embeddable watermarking capacity can be increased.

Error modeling method and device for prediction context of reversible image watermarking

Disclosed is a device for measuring speed of a tennis ball, which includes: a housing, a power supply, a main control device, a sensor device, a transmission device, a clamping device and a limiting device. The clamping device is clamped at the serve opening of a tennis ball machine, which can detect the serve speed of a tennis ball accurately with simple operation. The transmission device is engaged with the second sensor device, and the housing is slidably connected with the second sensor device. The emitting ends of the second sensor device and the first sensor device are positioned in the same vertical plane. The limiting device is connected with the second sensor device to limit excessive operation of the transmission device.

Device for measuring speed of tennis ball

A double attention network (DANet)-based drone patrol and inspection system for coastline floating garbage, including: an image acquisition module configured to shoot a video of a coastline in need of patrol and inspection by using a drone, and obtain an image from the video; a feature extraction module configured to extract shallow features and deep features, fuse the shallow features and the deep features to obtain a shared feature, and finally output a panoramic recognition result; a network training module configured to perform training on the labeled image so that the network can recognize the coastline and floating garbage; and a path correction module configured to adjust a flying direction of the drone.

DANet-based drone patrol and inspection system for coastline floating garbage

A method for measuring an antenna downtilt angle based on a deep instance segmentation network is disclosed, including: shooting an omni-directional antenna video using a drone; and transmitting the antenna video to a server in real time, and the server measuring an antenna downtilt angle in real time using a deep learning algorithm, the deep learning algorithm includes: a feature extraction network module for acquiring an antenna feature image; an instance segmentation module for binary segmentation of an antenna image and the background to distinguish antenna image pixels from background pixels; an antenna candidate box module for identifying and detecting the antenna image, acquiring an antenna candidate box and determining an antenna calibration box therefrom; and an antenna downtilt angle measuring module for measuring an antenna downtilt angle.

Method for measuring antenna downtilt angle based on deep instance segmentation network

Disclosed is a method for measuring an antenna downtilt based on a multi-scale detection algorithm, including: capturing an image of an antenna using an unmanned aerial vehicle, and returning data to a server in real time; obtaining a ground truth box where the antenna is located by performing the multi-scale detection algorithm of a server; segmenting the ground truth box based on an antenna target segmentation algorithm of the server; and obtaining an antenna downtilt angle based on an antenna downtilt measurement algorithm of the server and determining whether the antenna properly functions. This method avoids the danger of tower worker climbing, is fast and accurate, saves labor costs and time, and ensures the measurement of an antenna downtilt to be smoother.

Growing community of inventors

Jiangmen, China

Yikui Zhai