PSID-001 - Deepfake Video Detection
Design a machine learning model to detect deepfake videos with high precision, effectively distinguishing authentic videos from advanced forgeries. This model should assign a credibility score to each video, reflecting its likelihood of being genuine.
PSID-002 - Deepfake Audio Detection
Audio manipulation poses a growing threat. This hackathon challenges you to develop advanced ML models that can detect deepfakes in audio accurately and reliably & predict the authenticity of audio samples and allot a score indicating its credibility.
PSID-003 - Deepfake Text Detection
Create a ML model that exposes the subtle linguistic cues hidden in manipulated text, distinguishing real news from fabricated narratives. Develop an advanced text deepfake detection model that accurately predicts manipulated text.
PSID-004 - Deepfake Image Detection
Social media platforms are struggling to identify and remove harmful content. Your challenge is to develop an advanced ML model that analyzes images to detect deepfakes with unparalleled accuracy, even against state-of-the-art techniques.
Submission GuideLines
Use latest Python and Jupyter Notebooks or Google Colab for your machine learning model development environment.
Solutions must be submitted through a GitHub repository or a Google Drive link, containing all necessary files, including saved model weights.
Train your model on diverse, up-to-date datasets that include a variety of video qualities, formats, and both real and deepfake content. Ensure the inclusion of diverse ethnic samples.
Alongside your submission, provide a detailed report outlining the model development, training methodologies, and performance evaluation of your solution.