Rostom Ben Abdallah

Research intern working on an AI-powered computer vision pipeline for animal behavior analysis. The project focuses on detecting, tracking, and re-identifying animals in video data, then using time-series models to classify behavioral patterns. I work with large-scale video datasets, annotation files, object detection models, tracking algorithms, and HPC resources for model training and experiment

Title : AI-driven real-time tracking and counting for automated medication dispensing
• Collected on-site videos, annotated with polygons, and fine-tuned YOLOv11-seg to identify each
unit (tablets/vials).
• Built a real-time pipeline: detection → tracking (ByteTrack) → counting with OpenCV; generated
annotated videos and prevented double counting.
• Developed a PyQt operator UI (statuses, live fee

Title : Real Time Quality Control for lazer engraving using computer vision and AI
• Designed a real-time QC system for laser marking: industrial camera (3264×2448) + 45° mirror;
pipeline from acquisition → detection → validation → decision.
• Trained two specialized YOLOv11 models (words + anchor rectangle) and applied OpenCV to
verify position (px→mm), color (CIEDE2000 ΔE00), and opacity; used E

Title : AI-assisted poultry counting on a production line
• Collected and annotated thousands of images (Roboflow) and fine-tuned YOLOv8.
• Integrated the BoTSORT tracker and implemented IN/OUT counting logic (OpenCV); produced realtime annotated videos.
• Converted the model to ONNX, tested on Raspberry Pi 4 + camera v2, and optimized latency/FPS.
• Logged to MariaDB and generated productivity r

Worked as End of study internship at Tunisia Development Systems « TDS ».

I was a trainee at Industry 4.0 Digital Research Center Sfax from June 15 to july 1. During that time, I worked on several mini projects involving ESP32 microcontroller and a wide range of sensors, as well as on a project that involved creating and designing a PCB (printed circuit board)