Analyze an image or video
Upload an image or video in Streamlit. For video, the app samples frames for analysis.
Alameda Hacks / Python and Streamlit
A deepfake detector that explains why it flags an image and accepts user feedback.
The app combines machine learning with computer vision checks for signs of manipulation, including unusual compression patterns and inconsistent lighting. The Streamlit interface shows the evidence behind a verdict and lets users submit corrections. The project also supports sampling frames from video.
Upload an image or video in Streamlit. For video, the app samples frames for analysis.
The detector combines model predictions with forensic cues such as compression artifacts, unusual frequency patterns, and inconsistent lighting. The interface shows the evidence and confidence behind a result.
Users can mark a verdict as correct or wrong. The prototype uses that feedback to update its model and stores versioned checkpoints on Hugging Face.