Research Topics
We mainly focus on developing deep learning algorithms to solve many problems in the computer vision field. Including general tasks in computer vision such as classification or detection, we also study applicable work like generative models (GAN and diffusion models), depth estimation, medical imaging, noisy data, semantic segmentation, etc. Besides, there are many research on the fundamental structure of neural networks such as network minimization and architecture search in our laboratory. Our various research topics can be categorized into three sections which are Deep Learning Applications, Neural Network Architecture, and Learning Methods.
Deep Learning Applications
Our research topics in application of deep learning are focused on generative models (GAN, diffusion models), multi-modal learning, autonomous driving (depth estimation, object tracking, multi-sensor recognition), and anomaly detection (manufacturing, medical imaging).
Generative Models
Research focuses on improving various generative models such as StyleGAN and diffusion models to enhance image quality and provide smooth control over source features
Occlusion Removal
Removing or reconstructing occluded regions in an image, where an occlusion occurs when one object blocks another object from view
Multi-modal learning
Learning from multiple sources of data, such as vision-language. Aims to combine information from different modalities to improve performance and enhance understanding of the underlying data
Object Detection
Discover objects with directional orientation in bird's eye view images such as those captured by Radar or Satellite imaging
Anomaly Detection
Identifying unusual data points that deviate significantly from the normal or expected behavior in a dataset. Or, detecting abnormality in manufacturing process or medical images
Depth Estimation
Predicting the distance of objects in an image or video from a camera. It involves creating a model that can estimate the depth of each pixel in an image or a set of images, based on information such as the geometry of the scene and the movement of the camera
Neural Network Architecture
Our research on neural network architectures aims to find the optimal structure for convolutional neural networks. Recent topics include vision transformer, designing new architectures, or new operations in the network, and automated architecture search.
Lightweight Architecture
Designing a CNN-Transformer hybrid architecture with efficient cross-attention mechanism for lightweight RGB monocular depth estimation
Vision Transformer
Improving monocular depth estimation using a hierarchical transformer encoder, and lightweight decoder with selective feature fusion
Network Architecture
Designing a new CNN architecture with gradually increasing feature dimension (PyramidNet)
Neural Architecture Search
Automated architecture search with deep learning
Learning Methods
We study various topics in learning methods to solve problems in data used for training, such as noisy labels or biases. We also aim to improve the utilization of neural networks in real life through trustworthy ML, domain adaptation, and continual learning.
Noisy Labeled Data
Utilizing negative ground truth for robustness to noisy-labeled data (Negative Learning)
Biased Data
Learning useful information while not learning biases included in the data
Trustworthy ML
Research on vulnerabilities that exist in current deep learning models or systems (e.g., adversarial attack, model inversion attack, gradient inversion attack)
Domain Adaptation
Overcome the difference between the data distribution in the source domain and target domain and improve the model's performance on the target domain
Incremental Learning
Training a neural network with new data while keeping information learned from original data