About Me
I am Fedy Haj Ali, a Research Fellow at the AImageLab research group of the University of Modena and Reggio Emilia, under the supervision of Prof. Simone Calderara. My current focal point revolves around the intersection of Computer Vision and Deep Learning, specializing in the realms of 2D/3D object Detection and multiple object and people Tracking. In addition to my core focus, I am actively involved in a collaborative project with the Italian Institute IZS of Abruzzo and Molise related to AI application for identifying distress in animals during breeding and pre-slaughter phases.
For any questions or curiosity, feel free to reach out to me.
News
- [2023/09] I participated in the ELLIS Summer School on Large-Scale AI for Research and Industry in Modena, Italy.
- [2023/07] The paper TrackFlow: Multi-Object tracking with Normalizing Flows from our research group has been accepted at ICCV 2023. While I wasn't listed as an author, I'm proud to have been actively involved in this project for over four months, and I am grateful for the opportunity to have been part of such an innovative and impactful research effort.
- [2023/02] I started my collaboration with the Italian Institute IZS of Abruzzo and Molise.
- [2022/11] I joined the AImageLab group at UNIMORE in Italy as a Research Fellow.
Publications
DistFormer: Enhancing Local and Global Features for Monocular Per‑Object Distance Estimation
Aniello Panariello, Gianluca Mancusi, Fedy Haj Ali, Angelo Porrello, Simone Calderara, Rita Cucchiara
ArXiv Preprint, 2024
DistFormer, a revolutionary architecture for per-object distance estimation in safety-critical fields like surveillance or autonomous driving, outperforms existing methods with a robust context encoder, self-supervised encoder-decoder, and global refinement module. Extensive evaluations on KITTI, NuScenes, and MOTSynth datasets demonstrate its superior performance, resilience to challenges, and remarkable generalization, especially excelling in zero-shot synth-to-real transfer scenarios.
TrackFlow: Multi-Object Tracking with Normalizing Flows
Gianluca Mancusi, Aniello Panariello, Angelo Porrello, Matteo Fabbri, Simone Calderara, Rita Cucchiara
International Conference on Computer Vision (ICCV), 2023
We propose a novel approach to multi-object tracking that leverages Normalizing Flows to learn a joint probability distribution over the costs of candidate associations. Our experiments show that our approach consistently enhances the performance of several tracking-by-detection algorithms.
Projects
Security System Prototype
Fedy Haj Ali - 05/2021
The Security System project is a prototype for an alarm and security system that simulates the functionality of an anti-theft alarm by utilizing various hardware components. The project involves simulating the opening and closing of windows in an imaginary house using buttons and LEDs. Additionally, an ultrasonic distance sensor, specifically the HC SR04, is employed to detect any ‘suspicious’ movements within the simulated house. Upon detection, an acoustic signal is triggered to emit the alarm sound. To provide users with relevant information, an LCD display is incorporated to showcase messages and notifications. The system offers a user-friendly interface through a mobile application facilitated by MQTT (Message Queuing Telemetry Transport). This enables users to manage all functionalities connected with the security system, ensuring seamless communication between the hardware components and the mobile application, enhancing the overall efficiency and user experience of the prototype.
Academic Ticketing System
Fedy Haj Ali - 09/2021
Introducing our Academic Ticketing System, a user-friendly web app designed for efficient communication within academic institutions. Users can create tickets by selecting groups, topics, and receivers, with the flexibility to make changes throughout the resolution process. After resolution, tickets can be marked as validated or set to expire, ensuring timely response. The dedicated ticket section promotes seamless collaboration. Built with Django, Angular, Docker, redis, and PostgreSQL, our system offers a scalable, secure, and responsive solution for revolutionizing academic issue resolution.