Time Series
Time series models, which can predict or generate data in consideration of time-dependent variables, are one of the important topics in the field of machine learning. These models can help in many real-world tasks such as traffic forecasting and weather forecasting. We focus not only on the predictive performance of time-series models but also on improving the interpretability of the algorithms.
-
Interactive visual user
interfaces for automatic
sketch colorization -
User interaction allows users to generate or modify color images as intended while using automatic
colorization technique.
DAVIAN Lab. presented an exemplar-based sketch colorization that can effectively reflect color reference images provided by users with dense spatial correspondence.
Also, we research the method to enhance edges through user interactive edge hints and correct color bleeding artifacts in the generated images.
In addition, we promote a human-ai collaboration environment in which users can easily participate in processes by developing user interfaces.
- Air quality forecasting
-
Air quality forecasting is the application of science and technology to predict the composition of the air
pollution in the atmosphere for a given location and time.
DAVIAN Lab. attempt to solve the forecasting problem by applying attention based neural networks and graph neural networks. Additionally, to handle missing values inevitable in the sensor data, we also studying recurrent neural networks-based imputation techniques.