This team project detects whether Arabic text was written by a human or generated by AI. We prepared a balanced dataset of 12,796 samples and compared models including XGBoost and fine-tuned AraBERTv2. The best hybrid model reached 93.16% internal accuracy and 86% external accuracy.
Accuracy results of the hybrid Arabic text detection model.
This was a team project in the Samsung Innovation Campus AI Program. We used Python to prepare image data and trained YOLO models to locate and classify car damage across 27 classes. The selected YOLOv8 model achieved about 0.85 mAP@0.5 on the test set.
