FakeProfileDetection - Using Computer Vision Techniques to identify untrustworthy and fake profiles on various online platforms.
Clash Of Codes || Team Generating Adversarial Thoughts.
Theme: AI/ML.
Table of Contents
A feature rich API dedicated towards using Computer Vision techniques on input profile picture and identify potentially fake profiles.
We've focused on the following primary features :
- Provide different inference models to the users, so they can use the one with a balance of latency and accuracy as per their task.
- Identify false Positives : Detect cartoon/anime faces in the input images and distinguish them from human faces.
- Provide Additional details about the input image including Age, Gender, Potential Celebrity match, if its AI Generated, etc which might be relevant in fake profile detection.
Read the official documentation here
- Imporve the False positive detection model with a custom dataset : Current model is based on icartoon dataset which does not contain all types of cartoonish/animated face images. The model can be improved by a big margin if its trained on a better dataset.
- Improving the models for latency and providing support for wider variety of input images including sketches, DeepFakes, etc
- FastAPI
- OpenCV
- Scikit-Image
- TensorFlow
- dlib
- JavaScript
- React
- Tailwind