Spot the Deepfake
Can you spot the deepfake? Select which image you think is manipulated.
Which image is a deepfake?
Our cutting-edge AI system can identify manipulated videos with high accuracy
Our project addresses the growing threat of deepfake videos by leveraging the power of deep learning to detect manipulated media.
Deepfakes are synthetic media in which a person's likeness is replaced with someone else's using artificial intelligence. This technology can create convincing but fake videos of people saying or doing things they never did.
As deepfake technology becomes more accessible and sophisticated, it poses serious threats to information integrity, personal reputation, and public trust. Without reliable detection methods, these synthetic videos can be used for misinformation, fraud, or harassment.
DeepFake Insight provides a powerful, accessible tool that uses advanced neural networks to analyze videos and detect potential manipulation. Our system identifies subtle inconsistencies and artifacts that are invisible to the human eye.
Our system employs a sophisticated ResNeXt50 backbone combined with LSTM modules to analyze temporal inconsistencies across video frames. By examining both spatial and temporal features, our model achieves high accuracy in distinguishing between authentic and manipulated content.
Our model has been trained on diverse datasets of real and manipulated videos, achieving over 90% accuracy in controlled tests. The system analyzes facial inconsistencies, unnatural movements, and artifacts introduced during the synthesis process to deliver reliable results.
Can you spot the deepfake? Select which image you think is manipulated.
Our deepfake detection process involves multiple steps to ensure accurate analysis and results.
Start by uploading your video file through our secure interface. We accept common video formats including MP4, AVI, MOV, WMV, and MKV with a maximum size of 100MB.
Our system extracts key frames from the video and performs face detection to isolate facial regions, which are the primary focus of analysis for most deepfake manipulations.
The extracted frames are processed through our neural network architecture, which analyzes both spatial features within individual frames and temporal inconsistencies across the video sequence.
The system provides a clear verdict indicating whether the video is likely authentic or manipulated, along with a confidence score representing the reliability of the analysis.
Our detection system is built on a two-stage architecture:
$ docker pull bharshavardhanreddy924/deepfake_detection
The entire project is available as a Docker container, making it easy to deploy and integrate into your own applications.
Our system focuses on several key indicators of manipulation:
These indicators are weighted and analyzed collectively to produce a comprehensive assessment of the video's authenticity.
Upload a video and our AI will analyze it for signs of manipulation.
Our system analyzes facial videos to determine if they've been manipulated using deepfake technology. Upload a video to test it out.
Launch DemoSupported formats: MP4, AVI, MOV, WMV, MKV (max 100MB)
We're a group of students passionate about creating technology for good.
AIML, RV College of Engineering
AIML, RV College of Engineering
CSE, RV College of Engineering
CSE, RV College of Engineering
We leverage cutting-edge AI and computer vision techniques to deliver accurate deepfake detection.
Python
PyTorch
OpenCV
TensorFlow
Docker
Flask