We propose to create a virtual sandbox and then perform a series of operations on it through somatosensory interaction, including the 2 change of perspective, multi-directional observation, and the movement of war chess units. The whole project is divided into three parts: creation of a sandbox model, rendering of the virtual scene, and definition of somatosensory interaction.
Research on Cross-subject Emotion Recognition Model Based on Domain Generalization
To solve the accuracy loss caused by the individual differences of physiological signals, transfer learning has become a popular research direction in the cross-subject brain-computer interface system in recent years. Domain generalization is an algorithm in the direction of transfer learning in deep learning. This research is based on previous works, and is trying to optimize Domain Adversarial Neural Network (DANN) and Domain Residual Network (DResNet).
We propose to use average hashing and perceptual hashing algorithms, and image segmentation technology to preprocess the image pairing of the OCT medical image set. Then we implement a deep learning network model suitable for OCT medical image denoising based on a convolutional neural network and some other networks. Our innovation lies in the different types of images in the same data set, which is to perform supervised denoising for data sets with good pairing, and unsupervised denoising for data sets with insufficient pairing.
I implemented the compilation and programming of the single-chip program based on the Energia platform. Then I used Android Studio to develop a mobile phone APP, implementing intelligent car control driving based on Bluetooth communication. Finally, I implemented the functions of video return, key control, gesture recognition, gravity sensor, and voice control of an intelligent car driving.
Predictive Classification of Medical Images of Diabetic Retinopathy
To predict the extent of diabetic retinopathy, this project is mainly to learn about neural networks first, then reproduce previous people’s work. In detail, I established a neural network model based on the Keras CNN, and implemented the classification of diabetic retina 0-4 lesions through training and learning, then I built a neural network to realize the location and segmentation of the pathological area.
Campus Bus Query APP - IBUS
This project aims to provide a more convenient platform for students who need to take the school bus, avoiding wasting time waiting in line and ensuring teachers’ interests under the condition of the original priority.
