Hi, I'm Junsu 👋

Research Intern (Advisor: Prof. Ziwei Liu, NTU)

• Conducting research on 4D world modeling for embodied AI, integrating multi-view visual perception with video diffusion models and 3D scene representations.

Research Intern (Advisor: Prof. Seungryong Kim, KAIST AI)

• Conducting research on 3D vision: multi-view scene reconstruction pipelines integrating Gaussian Splatting and transformer-based depth estimation (VGGT / NoPoSplat).

Research Intern (Advisor: Prof. Mun Churl Kim, KAIST EE)

• Reviewed object detection models (RCNN–YOLO, ViT).
• Implemented baseline detectors for few-shot and anomaly detection tasks.

Two-stage metric-learning pipeline for face verification using ResNet-18 and AM-Softmax.

• Scratch training → fine-tuning structured training design
• Ablation on loss, optimizer, LR scheduler, and augmentation strategies
• Achieved 8.02% EER (baseline: 22.31%)