Sadegh Mohammadian

I am a fourth-year Computer Engineering student at Sharif University of Technology.

My research interests lie in computer vision, generative models, and multimodal learning.

Email  /  Google Scholar  /  Github  /  LinkedIn

Publications

Understanding Counting Mechanisms in Large Language and Vision-Language Models

Under Review, 2025

arXiv /

Uncovering Grounding IDs: How External Cues Shape Multi-Modal Binding

Under Review, 2025

arXiv /

Research Experiences

Sharif ML Lab
Vision Language Models • Apr. 2025 to Now
Advisor: Prof. Mahdieh Soleymani

Projects

Image Colorization

Implementing U-Net without skip connections using PyTorch for the image colorization task, trained and evaluated on the CIFAR-10 dataset to assess the effect of skip connections.

Code /

Mnist GAN

Implemented MLP discriminator and generator networks with PyTorch.

Trained and evaluated the results on the MNIST dataset and visualized the results at each step.

Code /

MobileNet V1 & V2

Implementing MobileNet V1 and V2 with PyTorch and comparing training and evaluation time with a normal CNN.

Code /

Semantic Segmentation Using U-Net

Applying a segmentation model to distinguish between different parts of the road, applicable in other downstream tasks such as self-driving vehicles. Using the Cityscapes dataset for training the model.

Code /

MLP Digit Classifier from Scratch

Implementing an MLP with NumPy from scratch to classify handwritten digits, trained and evaluated on the MNIST dataset.

Code /

Video Background Removal

Using SVD matrix factorization for image compression, background removal in videos, and foreground detection. NumPy was used for this project.

Code /

Design and source code from Jon Barron's website