DIO86342/Classification-of-Somali-Celebrity-Faces-Using-FaceNet-Embeddings-and-Support-Vector-Machine-SVM-

AI-powered Django web application for Somali celebrity face identification using MTCNN, FaceNet512, and SVM. Supports image upload and live camera capture with confidence scoring and biography display. # Somali Celebrity Face Identification System A Django-based web application that identifies Somali celebrities and public figures from an uploaded image or a live camera capture. The system integrates advanced computer vision pipelines with machine learning classifiers to detect faces, extract deep embeddings, verify identity, and present a full biographical summary to the user. --- ## 🚀 Key Features & Completed Steps * **Dual Input Modes:** Supports standard file uploads (`Choose File`) and dynamic live webcam streaming via the browser's Camera API. * **Seamless Form Integration:** Live camera captures are converted into JPEG Blobs/Files on the fly, feeding directly into the existing backend upload pipeline without breaking compatibility. * **Rigorous Verification:** Combines an SVM classifier with explicit cosine-distance metric checks against known database embeddings to prevent false positives. * **Rich Data Presentation:** Displays the processed image, predicted identity, model confidence score, and a comprehensive biographical record summary. * **Modern UI/UX:** Styled with a clean blue-and-white theme, completely mobile-responsive, featuring real-time local captured-image previews. * **Robust Error Handling:** Features empty-upload prevention, 0-byte capture filtering, and optimized in-memory image decoding using OpenCV. --- ## 🛠️ Technologies Used ### Backend & Core Logic * **Python** * **Django** (Web Framework & FileSystemStorage) * **SQLite** (Database) * **OpenCV (cv2)** & **NumPy** (Image decoding & processing) * **Pickle** (Model serialization & rapid loading) ### Machine Learning & Computer Vision * **MTCNN:** For highly accurate facial detection and localization. * **DeepFace (FaceNet512):** For extracting robust 512-dimensional facial embeddings. * **Scikit-Learn (SVM):** Main classification model mapping embeddings to celebrity labels. * **Cosine-Distance Metrics:** Secondary validation layer checking proximity to known face representations …