- Generative Adversarial Networks and Image Synthesis
- Image Enhancement Techniques
- Video Analysis and Summarization
- Text and Document Classification Technologies
- Topic Modeling
- Human Mobility and Location-Based Analysis
- Advanced Vision and Imaging
- Data Management and Algorithms
- Human Pose and Action Recognition
- Geographic Information Systems Studies
- Web Data Mining and Analysis
The University of Tokyo
2023
Dalian University of Technology
2021-2023
Automatic sketch colorization is a challenging task that aims to generate color image from sketch, primarily due its inherently ill-posed nature. While many approaches have shown promising results, two significant challenges remain: limited patterns and wide range of artifacts such as bleeding semantic inconsistencies among relevant regions. These issues stem the operation traditional convolutional structures, which capture structural features in pixel-wise manner, resulting inadequate...
Spatio-temporal compression of videos, utilizing networks such as Variational Autoencoders (VAE), plays a crucial role in OpenAI's SORA and numerous other video generative models. For instance, many LLM-like models learn the distribution discrete tokens derived from 3D VAEs within VQVAE framework, while most diffusion-based capture continuous latent extracted by 2D without quantization. The temporal is simply realized uniform frame sampling which results unsmooth motion between consecutive...
Muyao Niu, Jie Cai. Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and 9th International Joint (EMNLP-IJCNLP). 2019.
In our daily life, we prefer to describe the location of a person with semantic words such as "travel," "cinema" rather than geographic terminology GPS coordinates ($32^{\circ }24^{^{\prime }}12^{^{\prime \prime }}$32∘24'12''N, $2^{\circ }90^{^{\prime }}26.5^{^{\prime }}$2∘90'26.5''E). This is because places are easier understand compared precise location. Existing methods only take text-image pairs into consideration, neglecting user attribute, which essential in place prediction. this...