The AI Brain – Big Data and YOLOv11m Optimization

I. Dataset Curation under Ali KORKMAZ The primary force behind our AI's success is the meticulously prepared dataset, coordinated by Ali KORKMAZ and executed by the team members. Data Mining: 31,000 images tailored to competition scenarios were collected and labeled across various lighting angles, weather conditions, and altitudes. Data Augmentation: The dataset was enriched using techniques such as rotation, blurring, and brightness adjustments to ensure the model remains resilient against environmental variables. II. YOLOv11m Architecture and Training Process To achieve the optimal balance between speed and accuracy, we selected the YOLOv11m (Medium) version, one of the most advanced architectures available. Hardware & Optimization: Training was conducted on high-performance GPUs. Leveraging Mr. Korkmaz's Information Technology expertise, hyper-parameters were optimized to enhance the model's generalization capabilities while preventing overfitting. III. Performance Metrics: 91% mAP50 Success The validation tests conducted post-training yielded exciting results: Precision (mAP50): Our model achieved a success score of 0.91 (91%) in correctly identifying and classifying targets. Inference Speed: Thanks to the "m" version of YOLOv11, real-time and low-latency image processing capabilities were validated for our onboard computing system. IV. Continuous Improvement and Future Vision While the results are highly satisfactory, the team remains dedicated to excellence. Ali KORKMAZ and our students are continuing to refine the dataset to push accuracy toward 95%, specifically focusing on complex backgrounds and small-object detection.