Junling Zhou

Education

1.Significant Object Detection in Video Based on Graph Convolutional Network
Hangzhou,June 2023 - Present
Adivisor: Prof.Xiaofei Zhou,Hangzhou Dianzi University

Understanding traditional work practices: Understand the traditional methods of extracting spatiotemporal features using VGG or ResNet backbone networks, then use a certain strategy to fuse spatiotemporal features to generate deep features, and finally obtain significant prediction maps by decoding deep features
Improvement using graph convolution: Using graph convolution, Intra, and Inter GCNs methods for video salient object detection
Improve existing network structure: Using vit transformer instead of resnet50 to optimize network structure

Using 50 fold training to improve the model: Using the nnunet model based on PaddlePaddle and autonomously adding a 50 fold training method to achieve segmentation and transformation of medical images
3D imaging on web pages: Using Python’s Flask framework and Niivue’s front-end Javascript library, we deployed the unet model and implemented a web version of a 3D smart medical diagnosis platform

The front-end page displays a road prediction map: Build a path prediction and visualization website using the Vue framework and web development of HTML, CSS, and JS
Echarts chart visualization crawler collects data: Visualize the flow, composition, and pathway of people involved in trafficking using ECharts
Markov chain prediction probability: Using Markov chain prediction models to predict the probability of disappearance