Nahıl AutoML
End-to-end AutoML technique that designs task-specific preprocessing, weight initialization, architecture, activation & loss, and hyperparameters automatically
Same card language as the projects page with short summary + visual
End-to-end AutoML technique that designs task-specific preprocessing, weight initialization, architecture, activation & loss, and hyperparameters automatically
Supported by the Ministry of Health & TÜSEB. BIRADS classification on ~1000 radiologist-labeled mammograms with 90%+ success for mass & calcification detection
In production for banking/KYC: national ID OCR + MRZ, authenticity detection, and selfie–ID biometric face matching (IDOCR & IDBM)