How sak-cmd-web/face_recognition_system works
This is a webcam-based face recognition project with two separate recognition approaches: OpenCV LBPH and an InceptionResNetV2 classifier. The main workflow captures face images, trains a recognizer from labeled image folders, then detects and classifies faces from a live camera feed; LBPH also has a validation script. The README documents both detectors and both training/inference paths. Some referenced datasets and trained artifacts are absent from the tree, so their connections are shown as external data rather than repository nodes.
Main entry points: ml-lbph/webcam.py LBPH webcam recognition · train.py Train InceptionResNetV2 · dl-resnetv2/webcam.py DL webcam recognition · dnn_detector.py DNN SSD detector
Components 17
Image capture
- Capture faces
capture.py - Captured images
LBPH recognition
- Train LBPH
ml-lbph/train.py - LBPH model and labels
- Validate LBPH
ml-lbph/val.py - LBPH webcam recognition
ml-lbph/webcam.py
Deep learning
- Train InceptionResNetV2
dl-resnetv2/train.py - Class-name mapping
dl-resnetv2/utils.py - DL model and labels
- DL webcam recognition
dl-resnetv2/webcam.py
Face detection
- Haar detector
helpers/haar_detector.py - DNN SSD detector
helpers/dnn_detector.py
Other components
- User
- Webcam
- Labeled images
- SSD model files
source/deploy.prototxt - SSD weights
source/res10_300x300_ssd_iter_140000.caffemodel
Connections 22
- User → Capture faces: starts capture
- Capture faces → Webcam: reads frames
- Capture faces → Captured images: writes crops
- User → Train LBPH: starts training
- Labeled images → Train LBPH: provides training images
- Train LBPH → LBPH model and labels: writes model and labels
- Labeled images → Validate LBPH: provides validation images
- LBPH model and labels → Validate LBPH: loads model and labels
- User → LBPH webcam recognition: starts recognition
- Webcam → LBPH webcam recognition: supplies frames
- LBPH webcam recognition → LBPH model and labels: loads model and labels
- LBPH webcam recognition → DNN SSD detector: detects faces
- DNN SSD detector → SSD model files: loads configuration
- DNN SSD detector → SSD weights: loads weights
- User → Train InceptionResNetV2: starts training
- Labeled images → Train InceptionResNetV2: provides training and validation images
- Train InceptionResNetV2 → Class-name mapping: saves class mapping
- Train InceptionResNetV2 → DL model and labels: writes model and labels
- User → DL webcam recognition: starts recognition
- Webcam → DL webcam recognition: supplies frames
- DL webcam recognition → DL model and labels: loads model and labels
- DL webcam recognition → DNN SSD detector: detects faces