Fake-Face-Detector
- 0 Collaborators
Load and preprocess the dataset of fake and real images. Train the image classifier on the dataset of fake and real images. Use the trained classifier to predict if an image of a face was created by AI, or a photo taken of a real person. ...learn more
Project status: Published/In Market
Internet of Things, Artificial Intelligence
Intel Technologies
OpenVINO,
Intel Python
Overview / Usage
-
The goal of this project is to use a trained classifier to predict if an image of a face is real, or generated by AI.
-
The projects can aid in avoiding the use of any fake generated photos to create a fake ID.
-
The project can be integrated in production with any computerized photo upload to detect the photo's legitimacy.
Methodology / Approach
- Images were preprocessed(resized, cropped, and normalized) using pillow.
- Alexnet from torchvision.models pretrained models was used as the base model. It was loaded as a pre-trained network, based on which defined a new, untrained feed-forward network as a classifier, using ReLU activations and dropout. Trained the classifier layers using backpropagation using the pre-trained network to get the features. The loss and accuracy on the validation set were tracked to determine the best hyperparameters.
- Made predictions based on the trained model using openVino inference.
Technologies Used
- Python
- Pandas
- Numpy
- Matplotlib
- Pillow
- Requests
- Pytorch
- OpenVino