Javad Kavian
Computer Engineering Student
Welcome to my personal webpage. I am an undergraduate Computer Engineering student at the University of Tehran, with broad research interests in AI/ML, particularly in Computer Vision. I also work as a researcher at the RIML Lab at Sharif University of Technology.
Teaching Experience
During my undergraduate studies, I served as a teaching assistant for several undergraduate and graduate-level courses at the University of Tehran.
Courses
Deep Generative Models
GraduateMachine Learning
GraduateArtificial Intelligence
UndergraduateData Science
UndergraduateProgramming Languages and Compiler
UndergraduateIntroduction to Computing Systems and Programming
UndergraduateResearch Interests & Publications
My research interests are specifically focused on computer vision. In particular, I am interested in scene understanding and compositional visual reasoning — methods that can entirely understand intricate compositional patterns within an image.
Research Experience
Logical Anomaly Detection
Sharif University of Technology — RIML Lab
Scene Graph Generation
University of Tehran
Human-Computer Interaction (HCI)
University of Tehran
Publications
I currently have no published papers in peer-reviewed journals or conferences, but one of my works is under review at WACV 2027.
Projects
During my academic journey, I did several projects which you can find them on my github page →
Transformer from Scratch
A PyTorch implementation of transformer architecture, trained for translation task from english to persian
View on GitHub →Presentations
In our research journal clubs, I presented papers on recent advances in computer vision . You can find some of my presentation slides below.
Extracting Graph from Transformer for Scene Graph Generation
First Author: Jinbae Im · Im et al.
A scene graph generation model that efficiently extracts relational data by leveraging multi-head self-attention mechanisms already learned within a DETR-based object detector.
Grounding DINO: Marrying DINO with Grounded Pre-Training for Open-Set Object Detection
First Author: Shilong Liu · Liu et al.
An open-set object detector that bridges vision and language by integrating a transformer-based detector (DINO) with grounded pre-training, enabling the model to detect arbitrary objects based on text prompts.
Hybrid Reciprocal Transformer with Triplet Feature Alignment for Scene Graph Generation
First Author: De Cheng · Cheng et al.
A transformer-based scene graph generation method that combines triplet-level and component-level representations through reciprocal refinement and triplet feature alignment to improve predicate recognition in complex scenes with overlapping relationships.
Get in Touch
I am open to collaboration in research projects in the area of computer vision.
Email: javadkavian8@gmail.com
LinkedIn: /in/javadkavian
GitHub: github.com/javadkavian
Telegram: @javadmusiala