Chouchene Amrou

- Develop computer vision algorithms for American Football, focusing on players detection, players recognition, players tracking, and teams/players visual descriptors.

- Implement homography estimation techniques to map 3D field geometry onto 2D planes, enabling accurate players birds-eye view and enhanced gameplay visualization.

- Collaborate with cross-functional teams to integrate advanced AI/ML models into scalable platforms for sports analytics.

- Optimize algorithms for performance and scalability, ensuring high accuracy and reliability in dynamic environments.

Leveraged computer vision techniques to develop and implement cutting-edge solutions for detecting fraudulent actions in retail environments. Designed and deployed machine learning models and algorithms to identify suspicious behaviors, ensuring robust fraud prevention mechanisms. Key responsibilities included:

-Hand Classification: Developed deep learning models to classify hands as either containing objects or being empty, enabling the detection of concealed items during checkout processes.

- Product Segmentation: Used foreground-background separation to segment products with the lowest possible computational resources. Enhanced the precision of object boundaries, which improved downstream tasks like product identification and anomaly detection.

- Product Retrieval via Keypoint Matching: Engineered a system to compare product keypoints extracted from real-time images with a database of product images. Utilized feature extraction and matching techniques to retrieve corresponding products, enabling accurate verification of scanned items against actual purchases.

- Fraud Detection Optimization: Integrated classification, segmentation, and retrieval modules into a cohesive pipeline to detect fraudulent activities, including barcode swapping, item substitution, and mis-scanning. Collaborated with cross-functional teams to ensure seamless integration into existing retail systems, improving both security and operational efficiency.

- Performance Evaluation & Iteration: Conducted rigorous testing and evaluation of models using metrics such as precision, recall, and F1-score. Continuously iterated on model architectures and preprocessing pipelines to enhance accuracy and reduce false positives/negatives.

Worked on retail basket images captured by cameras to detect, segment, and recognize barcodes using advanced computer vision techniques and OCR (Optical Character Recognition). Leveraged cutting-edge algorithms to accurately isolate barcodes from complex backgrounds and ensure precise recognition of encoded information. This process involved preprocessing images, applying segmentation methods to extract barcode regions, and utilizing OCR tools to decode and validate the barcode data. The project aimed to enhance the efficiency and accuracy of automated retail systems, contributing to streamlined checkout processes and improved inventory management.