getting-started.md

Usage

Basic Usage

These set of instructions are for running the trained models.

Step 1:

Install all required dependencies by running (does not include tensorflow, opencv3): pip install -r requirements.txt

Step 2:

Place subjects as seperate folders under media/train_image_classifier.

There needs to be a minimum of two subjects.

Download the following models from here:

• SSD Mobilenet v1

• SSDLite mobilenet: Place under detection/model/ssdlite_v2.pb [This can be customised].

• Facenet Mobilenet: Place the frozen_inference.pb under detection/model/mobilenet_v2.pb [This can be customised].

• Facenet Inception_ Resnet_v1: Place the frozen_inference.pb under detection/model/inception_resnet_v1.pb [This can be customised].

Step 3:

Run python train_face_classify.py (mobilenet)

python train_face_classify.py --recognition_model inception_resnet_v1

For non-standard flags, run as:

python train_face_classify.py [-h]
                              [--recognition_model {inception_resnet_v1,mobilenet_v2,inception_resnet_v2}]
                              [--detection_model {ssdlite_v2,ssd_mobilenet}]
                              [--trained_classifier TRAINED_CLASSIFIER]
                              [--classifier {SVM,KNN}]
                              [--embedding_size {128,512}]

optional arguments:
  -h, --help            show this help message and exit
  --recognition_model {inception_resnet_v1,mobilenet_v2,inception_resnet_v2}
  --detection_model {ssdlite_v2,ssd_mobilenet}
                        detection model to use
  --trained_classifier TRAINED_CLASSIFIER
                        trained classifier to use
  --classifier {SVM,KNN}
                        trained classifier to use
  --embedding_size {128,512}
                        Embedding Size

Step 4:

Run

python demo_face_recognition.py (mobilenet)

python train_face_classify.py --recognition_model inception_resnet_v1 (inception_resnet)