RESEARCH
Deep learning-based Recognition
We are studying CNN-based object classification and detection robust to the environment changes such as illumination and pose changes
Deep Learning Network Architecture
We are studying deep learning network architectures for achieving better and best performances with less computational complexity
Domain Adaptation and Disentangle
We are studying Domain Adaptation and Disentangle methods for improving performances of deep learning methods without using a large size of training data and their corresponding ground truth
Bridging Large Gap of Domain Spaces for Unsupervised Domain Adaptation
Style and Content Disentangle with Autoencoder for classification
Domain Adaptation-based Labeling the human between real and cartoon images
KNOWLEDGE DISTILLATION
We are studying knowledge distillation methods for reducing the computational complexity of deep learning-based methods efficiently
On-the-fly Network Role Change for Online Knowledge Distillation
Pacemaker Knowledge Distillation for On-the-fly Convolutional Neural Network