How rizwann2912/deepfake-detection works
This Django application lets a user submit a video and sequence length for deepfake classification, then displays a REAL/FAKE result and confidence. The documented inference approach extracts ResNeXt features and classifies their sequence with an LSTM; the sampled view also shows face cropping and heatmap generation. Routing, upload validation, inference, and result presentation are visible, but the view excerpt omits parts of the request flow. The README describes model training, but no training implementation appears in the supplied tree.
Main entry points: views.py Request views · urls.py App routes · download_model.py Model downloader · templates Pages and results
Components 17
Web interface
- App routes
ml_app/urls.py - Upload form
ml_app/forms.py - Request views
ml_app/views.py - Pages and results
ml_app/templates - Shared page layout
templates - Upload preview script
static/js/script.js
Detection pipeline
- Frame and face processing
ml_app/views.py - Video sequence dataset
ml_app/views.py - ResNeXt LSTM classifier
ml_app/views.py - Prediction and heatmap
ml_app/views.py
Model assets
- Uploaded videos
uploaded_videos - Trained models
models - Model downloader
ml_app/management/commands/download_model.py
Django runtime
- Project routes
project_settings/urls.py - Application settings
project_settings/settings.py
Other components
- Video submitter
- Google Drive
Connections 16
- Video submitter → App routes: requests pages
ml_app/urls.py - Project routes → App routes: includes routes
project_settings/urls.py - App routes → Request views: dispatches requests
ml_app/urls.py - Request views → Upload form: validates input
ml_app/views.py - Request views → Uploaded videos: stores upload
- Application settings → Uploaded videos: configures media
project_settings/settings.py - Request views → Frame and face processing: processes video
ml_app/views.py - Request views → Video sequence dataset: loads sequences
ml_app/views.py - Request views → Trained models: loads weights
ml_app/views.py - Video sequence dataset → Frame and face processing: extracts frames
ml_app/views.py - Request views → ResNeXt LSTM classifier: runs inference
ml_app/views.py - Request views → Prediction and heatmap: generates result
ml_app/views.py - Request views → Pages and results: renders pages
ml_app/views.py - Pages and results → Shared page layout: uses layout
- Model downloader → Trained models: writes model
ml_app/management/commands/download_model.py - Model downloader → Google Drive: downloads from
ml_app/management/commands/download_model.py