Bilal Saeed Malik

Academic

Career

Verbatim text

UMASK (DEEP FAKE DETECTION SYSTEM)
Unmask is a web based deep fake video detection system designed to detect the forged and real videos. Here the users can provide the image or a video to the model and it classifies it as real or fake. The number of frames can be selected by the user for the given input. To get an overall view of how deep fakes work we give an option to the user of creating a deep fake video too, for creating a deep fake video the face image which needs to be swapped should be provided and it will be automatically swapped with the real face present in the video.
The web app features include:
- The subscription plans are provided to the user to use the premium features. Premium features cost may vary according to the length and fps rate of the video provided.
- Realtime results are provided for the videos with the confidence percentage in predicting the results.
- Feature Maps are generated by the system to give an overall view of how the system ended up in getting the results.
- Profile of users are maintained where they can get analysis of their activity on our website and the subscription model they have taken.
UNMASK
Detecting Deep Fake videos as real or forged
Iterations
1 FACE DETECTION AND CROPPING
2 FEATURE EXTRACTION (RES NEXT)
3 VIDEO CLASSIFICATION (LSTM)
4 MODEL INTEGRATION AND WEB APPLICATION
WORK FLOW
Data Collection Spliting Data
Uploading Video Spliting Video in to frames Pre-processing
Loading Trained Model Model Export Model Training
Prediction Real/Forge
Team Members:
19I-2016 Bilal Saeed Malik
19I-0538 Zain Ul Abidin
19I-2155 Bilal Ali
Supervisor: Mr. Shams Farooq
Co-supervisor: Mr. Bilal Khalid Dar
Technology Stack
OpenCV Keras JS
Technology Used:
Python, Pytorch, TensorFlow, React, Node, Next Js, Docker, Kubernets
Supervisor Name:
Mr. Shams Farooq
Group Members:
Bilal Saeed Malik (19I - 2016)
Zain Ul Abidin (19I - 0538)
Bilal Ali (19I - 2155)

AI enrichment