|
I am a senior Bachelor of Science student in Computer Engineering specializing in Artificial Intelligence at Amirkabir University of Technology. I am currently a research assistant in the SML lab under the supervision of Prof. Nickabadi, engaged in several research projects focused on visual place recognition. Additionally, I am collaborating with Dr. Alireza Esmaeilzehi on projects related to image super-resolution and generative models. Previously, I completed an internship at IPM, where I worked with Dr. Mohammad Sabokrou on enhancing model interpretability and robustness. |
|
|
My research interests broadly include machine learning and computer vision. Specifically, I focus on areas such as image/video restoration (e.g., super-resolution), Human Pose Estimation (HPE), image/video recognition, and robustness. Additionally, I am exploring the efficiency aspect of vision transformers, like token reduction, within these domains. |
|
|
|
Mohammad Javad Rajabi, Morteza Mirzai, Ahmad Nickabadi |
|
Alireza Esmaeilzehi, Mohammad Javad Rajabi, M. Omair Ahmad |
|
Raha Ahmadi*, Mohammad Javad Rajabi*, Mohammad Khalooie, Mohammad Sabokrou ACML 2023 |
|
|
| Jan 2024 |
Pose-and-Image Guided Diffusion-based Video Synthesis Models Journal Club [Slides] |
|---|---|
| Dec 2023 |
More Efficient Vision Transformers By Token Reduction AAISS 2023 [Slides] |
| Oct 2023 |
Token Merging Journal Club [Slides] |
|
|
|
Supervisor: Alireza Esmaeilzehi (Postdoctoral Fellow at UofT) Engaged in a research collaboration working on two innovative projects in video and image processing. The first is a video-to-video translation project aimed at improving temporal and spatial consistency using optical flow techniques. The second involves developing an image super-resolution network that uses a transformers-based approach with correlated cross-attention architecture. |
|
Supervisor: Ahmad Nickabadi Worked on Landmark Detection (Visual Place Recognition). Our primary focus was to enhance the accuracy of landmark image recognition by minimizing the impact of occluding objects, with a particular emphasis on human presence. By mitigating the influence of occluding objects, such as humans, we successfully improved the overall precision and reliability of the landmark recognition system. |
|
Supervisor: Mohammad Sabokrou Proposed a novel approach that encourages the model to focus on the primary object of the image and mitigates biases in image classification. This approach leads to an improvement in classification accuracy, reduced sensitivity to background noise, and enhanced model interpretability. |
|
|
| Spring 2023 |
Teaching Assistant for Information Retrieval Under the supervision of Dr. Ahmad Nikabadi |
|---|---|
| Spring 2023 |
Teaching Assistant for Robotics Under the supervision of Dr. Mahdi Javanmardi |
| Spring 2023 |
Teaching Assistant for Principles of Cloud Computing Under the supervision of Dr. S.Ahmad Javadi |
| Fall 2022 |
Teaching Assistant for Information Retrieval Under the supervision of Dr. Ahmad Nikabadi |
| Fall 2021 |
Teaching Assistant for Algorithm Design Under the supervision of Dr. Alireza Bagheri |
| Fall 2021 |
Teaching Assistant for Logic Circuits Under the supervision of Dr. Morteza Saheb Zamani |
| Fall 2020 |
Teaching Assistant for Fundamentals of Computer Programming Under the supervision of Dr. Hossein Zeinali |
|
|
 |
|
 |
|
|
Template adapted from Jon Barron's website. |