Research Focus
AIIU Lab studies visual intelligence across perception, generation, robustness, biometrics, tracking, 3D reasoning, multimodal reasoning, and autonomous validation.
Lab Direction
Our work connects foundation models with real visual data, physical scenes, and reliable evaluation for practical AI systems.
Research Areas
AI Security
Watermark robustness, adversarial attacks, backdoors, model safety, and trustworthy visual generation.
Deepfake Detection
Foundation-model adaptation, explainable detection, face forgery analysis, and robust media forensics.
MLLM
Vision-language reasoning, hallucination analysis, multimodal evaluation, and reliable foundation models.
Physical AI
Object pose, 3D/4D scene understanding, stereo geometry, robotic perception, and real-world visual grounding.
Agentic AI
LLM/VLLM agents for data generation, validation, prompt optimization, and autonomous visual workflows.
Generative Vision
Controllable diffusion, image/video synthesis, restoration, concept control, and creative visual modeling.
Recent News
View all publications2026
New papers are to appear in ICIP, ICML, CVPR Findings, FG, and ICLR.
Mar 2026
Four papers will appear at WACV 2026, spanning diffusion control, object pose estimation, domain-adaptive detection, and concept erasure evaluation.
Dec 2025
New work accepted to NeurIPS 2025 and ACML 2025.
Sep 2025
A strong run of papers appeared at ICIP 2025, MMSP 2025, and AVSS 2025, including oral and spotlight presentations.
Apr-Jul 2025
AIIU-related work appeared at ICLR 2025, ICML 2025, and CVPR 2025.
Featured Works
CVPR Findings 2026
Gen-n-Val: Agentic Image Data Generation and Validation
An agentic pipeline for generating and validating synthetic visual data for robust computer vision training.