FaceMatch: AI-Driven Forensic Sketching and Identification

FaceMatch: An AI-Driven Forensic Facial Composite and Identification System Using Deep Learning. FaceMatch, an AI-driven forensic facial composite and identification system for law enforcement. The system enables creating composite sketches and matching them against large photo databases using deep learning.FaceMatch uses a hybrid face recognition approach combining the `face_recognition` library (dlib- based ResNet-34) with an OpenCV fallback. The primary method achieves 99.38% accuracy on standard benchmarks using 128-dimensional face encodings and cosine similarity. The system is trained on 20,655 face images and tested on 679 images, with pre-built indexing for fast retrieval. The architecture includes a React/TypeScript frontend with a drag-and-drop composite sketch builder and a Flask REST API backend. The backend uses vectorized similarity calculations, parallel processing, and caching to reduce comparison time from minutes to seconds. The system supports sketch-to-photo matching by comparing user-created sketches against a pre-indexed sketch database and returning corresponding photos.