Model to recognize celebrities using a face matching algorithm.
Refer this for detailed documentation.
You can also read my article on medium here.
Basic working of the algorithm includes the following:
Face detection is done using MTCNN face detection model.
Face encodings are created using VGGFace model in keras.
Face matching is done using annoy library (spotify).
pip install -r requirements.txtto install all the dependencies (preferably in a virtual environment).
- To ensure you have all the required additional packages, run
pip install -r requirements.txtfirst.
- To install pip package, run:
# pip release version pip install celeb-detector # also install additional dependencies with this (if not installed via requirements.txt file) pip install annoy keras-vggface keras-applications # Directly from repo pip install git+https://github.com/shobhit9618/celeb_recognition.git
- If you are using conda on linux or ubuntu, you can use the following commands to create and use a new environment called celeb-detector:
This will install all the required dependencies. To ensure you are using the latest version of the package, also run (inside the environment):
conda env create shobhit9618/celeb-detector conda activate celeb-detector
pip install --upgrade celeb-detector
Using pip pakcage
For using my model for predictions, use the following lines of code after installation:
import celeb_detector # on running for the first time, this will download vggface model img_path = 'sample_image.jpg' celeb_detector.celeb_recognition(img_path) # on running for the first time, 2 files (celeb_mapping.json and celeb_index_60.ann) will downloaded to the home directory # if you want to use an image url, just provide the url and add url=True url = 'https://sample/sample_image_url.jpg' celeb_detector.celeb_recognition(url, url=True)
This returns a list of dictionaries, each dictionary contains bbox coordinates, celeb name and confidence for each face detected in the image (celeb name will be unknown if no matching face detected).
For using your own custom model, also provide path to json and ann files as shown below:
import celeb_detector img_path = 'sample_image.jpg' ann_path = 'sample_index.ann' celeb_map = 'sample_mapping.json' celeb_detector.celeb_recognition(img_path, ann_path, celeb_map)
For creating your own model (refer this for more details on usage) and run as follows:
import celeb_detector folder_path = 'celeb_images' celeb_detector.create_celeb_model(folder_path)
Create your own celeb model
- Create a dataset of celebs in the following directory structure:
celeb_images/ celeb-a/ celeb-a_1.jpg celeb-a_2.jpg ... celeb-b/ celeb-b_1.jpg celeb-b_1.jpg ... ...
- Each folder name will be considered as the corresponding celeb name for the model (WARNING: Do not provide any special characters or spaces in the names).
- Make sure each image has only 1 face (of the desired celebrity), if there are multiple faces, only the first detected face will be considered.
- Provide path to the dataset folder (for example,
celeb_imagesfolder) in the create_celeb_model.py file.
- Run create_celeb_model.py file.
- Upon successful completion of the code, we get
celeb_mapping.json(for storing indexes vs celeb names),
celeb_index.ann(ann file for searching encodings) and
celeb_name_encoding.pklfiles (for storing encodings vs indexes for each celeb). (WARNING: You need to provide paths for storing each of these files, default is to store in the current directory)
Model predictions in jupyter
- Provide paths to
celeb_index.annfiles in celeb_recognition.ipynb file. If you want to try my model, ignore this step.
- Run all the cells in the celeb_recognition.ipynb file, the final cell will provide widgets for uploading images and making predictions (this will also download the necessary model files).
- NOTE: celeb_recognition.ipynb is a standalone file and does not require any other files from the repo for running.
Model predictions in python
- Provide paths to
celeb_index.annfiles in celeb_recognition.py and celeb_utils.py files. If you want to try my model, ignore this step.
- Run celeb_recognition.py file, provide path to image in the file.
- Output includes a list of the identified faces, bounding boxes and the predicted celeb name (unknown if not found).
- It also displays the output with bounding boxes.
Sample image output
You can run a binder application by clicking the following link:
You can also launch a voila binder application (which only has widgets for image upload and celeb prediction) by clicking here.
To open and run celeb_recognition.ipynb file in google colab, click the following link: